def testAddNoise(self):
patternMachine = PatternMachine(10000, 1000, num=1)
pattern = patternMachine.get(0)
noisy = patternMachine.addNoise(pattern, 0.0)
self.assertEqual(len(pattern & noisy), 1000)
noisy = patternMachine.addNoise(pattern, 0.5)
self.assertTrue(400 < len(pattern & noisy) < 600)
noisy = patternMachine.addNoise(pattern, 1.0)
self.assertTrue(50 < len(pattern & noisy) < 150)
def testWriteRead(self):
tm1 = TemporalMemory(
columnDimensions=(100,),
cellsPerColumn=4,
activationThreshold=7,
initialPermanence=0.37,
connectedPermanence=0.58,
minThreshold=4,
maxNewSynapseCount=18,
permanenceIncrement=0.23,
permanenceDecrement=0.08,
seed=91
)
# Run some data through before serializing
patternMachine = PatternMachine(100, 4)
sequenceMachine = SequenceMachine(patternMachine)
sequence = sequenceMachine.generateFromNumbers(range(5))
for _ in range(3):
for pattern in sequence:
tm1.compute(pattern)
proto1 = TemporalMemoryProto_capnp.TemporalMemoryProto.new_message()
tm1.write(proto1)
# Write the proto to a temp file and read it back into a new proto
with tempfile.TemporaryFile() as f:
proto1.write(f)
f.seek(0)
proto2 = TemporalMemoryProto_capnp.TemporalMemoryProto.read(f)
# Load the deserialized proto
tm2 = TemporalMemory.read(proto2)
# Check that the two temporal memory objects have the same attributes
self.assertEqual(tm1, tm2)
# Run a couple records through after deserializing and check results match
tm1.compute(patternMachine.get(0))
tm2.compute(patternMachine.get(0))
self.assertEqual(set(tm1.getActiveCells()), set(tm2.getActiveCells()))
self.assertEqual(set(tm1.getPredictiveCells()),
set(tm2.getPredictiveCells()))
self.assertEqual(set(tm1.getWinnerCells()), set(tm2.getWinnerCells()))
self.assertEqual(tm1.connections, tm2.connections)
tm1.compute(patternMachine.get(3))
tm2.compute(patternMachine.get(3))
self.assertEqual(set(tm1.getActiveCells()), set(tm2.getActiveCells()))
self.assertEqual(set(tm1.getPredictiveCells()),
set(tm2.getPredictiveCells()))
self.assertEqual(set(tm1.getWinnerCells()), set(tm2.getWinnerCells()))
self.assertEqual(tm1.connections, tm2.connections)
def testWList(self):
w = [4, 7, 11]
patternMachine = PatternMachine(100, w, num=50)
widths = dict((el, 0) for el in w)
for i in range(50):
pattern = patternMachine.get(i)
width = len(pattern)
self.assertTrue(width in w)
widths[len(pattern)] += 1
for i in w:
self.assertTrue(widths[i] > 0)
def testWList(self):
w = [4, 7, 11]
patternMachine = PatternMachine(100, w, num=50)
widths = dict((el, 0) for el in w)
for i in range(50):
pattern = patternMachine.get(i)
width = len(pattern)
self.assertTrue(width in w)
widths[len(pattern)] += 1
for i in w:
self.assertTrue(widths[i] > 0)
def testWrite(self):
tm1 = TemporalMemory(
columnDimensions=[100],
cellsPerColumn=4,
activationThreshold=7,
initialPermanence=0.37,
connectedPermanence=0.58,
minThreshold=4,
maxNewSynapseCount=18,
permanenceIncrement=0.23,
permanenceDecrement=0.08,
seed=91
)
# Run some data through before serializing
self.patternMachine = PatternMachine(100, 4)
self.sequenceMachine = SequenceMachine(self.patternMachine)
sequence = self.sequenceMachine.generateFromNumbers(range(5))
for _ in range(3):
for pattern in sequence:
tm1.compute(pattern)
proto1 = TemporalMemoryProto_capnp.TemporalMemoryProto.new_message()
tm1.write(proto1)
# Write the proto to a temp file and read it back into a new proto
with tempfile.TemporaryFile() as f:
proto1.write(f)
f.seek(0)
proto2 = TemporalMemoryProto_capnp.TemporalMemoryProto.read(f)
# Load the deserialized proto
tm2 = TemporalMemory.read(proto2)
# Check that the two temporal memory objects have the same attributes
self.assertEqual(tm1, tm2)
# Run a couple records through after deserializing and check results match
tm1.compute(self.patternMachine.get(0))
tm2.compute(self.patternMachine.get(0))
self.assertEqual(tm1.activeCells, tm2.activeCells)
self.assertEqual(tm1.predictiveCells, tm2.predictiveCells)
self.assertEqual(tm1.winnerCells, tm2.winnerCells)
self.assertEqual(tm1.connections, tm2.connections)
tm1.compute(self.patternMachine.get(3))
tm2.compute(self.patternMachine.get(3))
self.assertEqual(tm1.activeCells, tm2.activeCells)
self.assertEqual(tm1.predictiveCells, tm2.predictiveCells)
self.assertEqual(tm1.winnerCells, tm2.winnerCells)
self.assertEqual(tm1.connections, tm2.connections)
def setUp(self):
print ("\n"
"======================================================\n"
"Test: {0} \n"
"{1}\n"
"======================================================\n"
).format(self.id(), self.shortDescription())
self.patternMachine = PatternMachine(1024, 20, num=100)
self.sequenceMachine = SequenceMachine(self.patternMachine)
self.tm = MonitoredTemporalMemory(
mmName="TM",
columnDimensions=[1024],
cellsPerColumn=16,
initialPermanence=0.5,
connectedPermanence=0.7,
minThreshold=20,
maxNewSynapseCount=30,
permanenceIncrement=0.1,
permanenceDecrement=0.02,
activationThreshold=20)
self.tp = MonitoredTemporalPooler(
inputDimensions=[1024, 16],
columnDimensions=[1024],
mmName="TP")
self._showInternalStatePlots(appendTitle=" (initial)")
def testAddSpatialNoise(self):
patternMachine = PatternMachine(10000, 1000, num=100)
sequenceMachine = SequenceMachine(patternMachine)
numbers = range(0, 100)
numbers.append(None)
sequence = sequenceMachine.generateFromNumbers(numbers)
noisy = sequenceMachine.addSpatialNoise(sequence, 0.5)
overlap = len(noisy[0] & patternMachine.get(0))
self.assertTrue(400 < overlap < 600)
sequence = sequenceMachine.generateFromNumbers(numbers)
noisy = sequenceMachine.addSpatialNoise(sequence, 0.0)
overlap = len(noisy[0] & patternMachine.get(0))
self.assertEqual(overlap, 1000)
def testAddSpatialNoise(self):
patternMachine = PatternMachine(10000, 1000, num=100)
sequenceMachine = SequenceMachine(patternMachine)
numbers = range(0, 100)
numbers.append(None)
sequence = sequenceMachine.generateFromNumbers(numbers)
noisy = sequenceMachine.addSpatialNoise(sequence, 0.5)
overlap = len(noisy[0] & patternMachine.get(0))
self.assertTrue(400 < overlap < 600)
sequence = sequenceMachine.generateFromNumbers(numbers)
noisy = sequenceMachine.addSpatialNoise(sequence, 0.0)
overlap = len(noisy[0] & patternMachine.get(0))
self.assertEqual(overlap, 1000)
def setUp(self):
"""
Sets up the test.
"""
# single column case
self.pooler = None
# multi column case
self.poolers = []
# create pattern machine
self.proximalPatternMachine = PatternMachine(
n=self.inputWidth,
w=self.numOutputActiveBits,
num=200,
seed=self.seed)
self.patternId = 0
np.random.seed(self.seed)
def generateSequences(patternDimensionality, patternCardinality,
sequenceLength, sequenceCount):
patternAlphabetSize = sequenceLength * sequenceCount
patternMachine = PatternMachine(patternDimensionality, patternCardinality,
patternAlphabetSize)
sequenceMachine = SequenceMachine(patternMachine)
numbers = sequenceMachine.generateNumbers(sequenceCount, sequenceLength)
generatedSequences = sequenceMachine.generateFromNumbers(numbers)
return generatedSequences
class PatternMachineTest(unittest.TestCase):
def setUp(self):
self.patternMachine = PatternMachine(10000, 5, num=50)
def testGet(self):
patternA = self.patternMachine.get(48)
self.assertEqual(len(patternA), 5)
patternB = self.patternMachine.get(49)
self.assertEqual(len(patternB), 5)
self.assertEqual(patternA & patternB, set())
def testGetOutOfBounds(self):
args = [50]
self.assertRaises(IndexError, self.patternMachine.get, *args)
def testAddNoise(self):
patternMachine = PatternMachine(10000, 1000, num=1)
pattern = patternMachine.get(0)
noisy = patternMachine.addNoise(pattern, 0.0)
self.assertEqual(len(pattern & noisy), 1000)
noisy = patternMachine.addNoise(pattern, 0.5)
self.assertTrue(400 < len(pattern & noisy) < 600)
noisy = patternMachine.addNoise(pattern, 1.0)
self.assertTrue(50 < len(pattern & noisy) < 150)
def testNumbersForBit(self):
pattern = self.patternMachine.get(49)
for bit in pattern:
self.assertEqual(self.patternMachine.numbersForBit(bit), set([49]))
def testNumbersForBitOutOfBounds(self):
args = [10000]
self.assertRaises(IndexError, self.patternMachine.numbersForBit, *args)
def testNumberMapForBits(self):
pattern = self.patternMachine.get(49)
numberMap = self.patternMachine.numberMapForBits(pattern)
self.assertEqual(numberMap.keys(), [49])
self.assertEqual(numberMap[49], pattern)
def testWList(self):
w = [4, 7, 11]
patternMachine = PatternMachine(100, w, num=50)
widths = dict((el, 0) for el in w)
for i in range(50):
pattern = patternMachine.get(i)
width = len(pattern)
self.assertTrue(width in w)
widths[len(pattern)] += 1
for i in w:
self.assertTrue(widths[i] > 0)
def setUp(self):
"""
Sets up the test.
"""
self.pooler = None
self.proximalPatternMachine = PatternMachine(
n=self.inputWidth,
w=self.numOutputActiveBits,
num=200,
seed=self.seed
)
self.patternId = 0
np.random.seed(self.seed)
def setUp(self):
self.tmPy = TemporalMemoryPy(columnDimensions=[2048],
cellsPerColumn=32,
initialPermanence=0.5,
connectedPermanence=0.8,
minThreshold=10,
maxNewSynapseCount=12,
permanenceIncrement=0.1,
permanenceDecrement=0.05,
activationThreshold=15)
self.tmCPP = TemporalMemoryCPP(columnDimensions=[2048],
cellsPerColumn=32,
initialPermanence=0.5,
connectedPermanence=0.8,
minThreshold=10,
maxNewSynapseCount=12,
permanenceIncrement=0.1,
permanenceDecrement=0.05,
activationThreshold=15)
self.tp = TP(numberOfCols=2048,
cellsPerColumn=32,
initialPerm=0.5,
connectedPerm=0.8,
minThreshold=10,
newSynapseCount=12,
permanenceInc=0.1,
permanenceDec=0.05,
activationThreshold=15,
globalDecay=0,
burnIn=1,
checkSynapseConsistency=False,
pamLength=1)
self.tp10x2 = TP10X2(numberOfCols=2048,
cellsPerColumn=32,
initialPerm=0.5,
connectedPerm=0.8,
minThreshold=10,
newSynapseCount=12,
permanenceInc=0.1,
permanenceDec=0.05,
activationThreshold=15,
globalDecay=0,
burnIn=1,
checkSynapseConsistency=False,
pamLength=1)
self.patternMachine = PatternMachine(2048, 40, num=100)
self.sequenceMachine = SequenceMachine(self.patternMachine)
def generateSequences(n=2048, w=40, sequenceLength=5, sequenceCount=2,
sharedRange=None, seed=42):
"""
Generate high order sequences using SequenceMachine
"""
# Lots of room for noise sdrs
patternAlphabetSize = 10*(sequenceLength * sequenceCount)
patternMachine = PatternMachine(n, w, patternAlphabetSize, seed)
sequenceMachine = SequenceMachine(patternMachine, seed)
numbers = sequenceMachine.generateNumbers(sequenceCount, sequenceLength,
sharedRange=sharedRange )
generatedSequences = sequenceMachine.generateFromNumbers(numbers)
return sequenceMachine, generatedSequences, numbers
def testAddNoise(self):
patternMachine = PatternMachine(10000, 1000, num=1)
pattern = patternMachine.get(0)
noisy = patternMachine.addNoise(pattern, 0.0)
self.assertEqual(len(pattern & noisy), 1000)
noisy = patternMachine.addNoise(pattern, 0.5)
self.assertTrue(400 < len(pattern & noisy) < 600)
noisy = patternMachine.addNoise(pattern, 1.0)
self.assertTrue(50 < len(pattern & noisy) < 150)
def setUp(self):
"""
Sets up the test.
"""
# single column case
self.pooler = None
# multi column case
self.poolers = []
# create pattern machine
self.proximalPatternMachine = PatternMachine(
n=self.inputWidth,
w=self.numOutputActiveBits,
num=200,
seed=self.seed
)
self.patternId = 0
np.random.seed(self.seed)
def generateSequences(patternDimensionality, patternCardinality, sequenceLength,
sequenceCount):
# Generate a sequence list and an associated labeled list
# (both containing a set of sequences separated by None)
print "Generating sequences..."
patternAlphabetSize = sequenceLength * sequenceCount
patternMachine = PatternMachine(patternDimensionality, patternCardinality,
patternAlphabetSize)
sequenceMachine = SequenceMachine(patternMachine)
numbers = sequenceMachine.generateNumbers(sequenceCount, sequenceLength)
generatedSequences = sequenceMachine.generateFromNumbers(numbers)
sequenceLabels = [
str(numbers[i + i * sequenceLength: i + (i + 1) * sequenceLength])
for i in xrange(sequenceCount)]
labeledSequences = []
for label in sequenceLabels:
for _ in xrange(sequenceLength):
labeledSequences.append(label)
labeledSequences.append(None)
return generatedSequences, labeledSequences
def generateSequences(patternCardinality, patternDimensionality,
numberOfSequences, sequenceLength, consoleVerbosity):
patternAlphabetSize = sequenceLength * numberOfSequences
patternMachine = PatternMachine(patternDimensionality, patternCardinality,
patternAlphabetSize)
sequenceMachine = SequenceMachine(patternMachine)
numbers = sequenceMachine.generateNumbers(numberOfSequences,
sequenceLength)
inputSequences = sequenceMachine.generateFromNumbers(numbers)
inputCategories = []
for i in xrange(numberOfSequences):
for _ in xrange(sequenceLength):
inputCategories.append(i)
inputCategories.append(None)
if consoleVerbosity > 1:
for i in xrange(len(inputSequences)):
if inputSequences[i] is None:
print
else:
print "{0} {1}".format(inputSequences[i], inputCategories[i])
return inputSequences, inputCategories
class ExtensiveColumnPoolerTest(unittest.TestCase):
"""
Algorithmic tests for the ColumnPooler region.
Each test actually tests multiple aspects of the algorithm. For more
atomic tests refer to column_pooler_unit_test.
The notation for objects is the following:
object{patternA, patternB, ...}
In these tests, the proximally-fed SDR's are simulated as unique (location,
feature) pairs regardless of actual locations and features, unless stated
otherwise.
"""
inputWidth = 2048 * 8
numInputActiveBits = int(0.02 * inputWidth)
outputWidth = 4096
numOutputActiveBits = 40
seed = 42
def testNewInputs(self):
"""
Checks that the behavior is correct when facing unseed inputs.
"""
self.init()
# feed the first input, a random SDR should be generated
initialPattern = self.generateObject(1)
self.learn(initialPattern, numRepetitions=1, newObject=True)
representation = self._getActiveRepresentation()
self.assertEqual(
len(representation),
self.numOutputActiveBits,
"The generated representation is incorrect"
)
# feed a new input for the same object, the previous SDR should persist
newPattern = self.generateObject(1)
self.learn(newPattern, numRepetitions=1, newObject=False)
newRepresentation = self._getActiveRepresentation()
self.assertNotEqual(initialPattern, newPattern)
self.assertEqual(
newRepresentation,
representation,
"The SDR did not persist when learning the same object"
)
# without sensory input, the SDR should persist as well
emptyPattern = [set()]
self.learn(emptyPattern, numRepetitions=1, newObject=False)
newRepresentation = self._getActiveRepresentation()
self.assertEqual(
newRepresentation,
representation,
"The SDR did not persist after an empty input."
)
def testLearnSinglePattern(self):
"""
A single pattern is learnt for a single object.
Objects: A{X, Y}
"""
self.init()
object = self.generateObject(1)
self.learn(object, numRepetitions=2, newObject=True)
# check that the active representation is sparse
representation = self._getActiveRepresentation()
self.assertEqual(
len(representation),
self.numOutputActiveBits,
"The generated representation is incorrect"
)
# check that the pattern was correctly learnt
self.pooler.reset()
self.infer(feedforwardPattern=object[0])
self.assertEqual(
self._getActiveRepresentation(),
representation,
"The pooled representation is not stable"
)
# present new pattern for same object
# it should be mapped to the same representation
newPattern = [self.generatePattern()]
self.learn(newPattern, numRepetitions=2, newObject=False)
# check that the active representation is sparse
newRepresentation = self._getActiveRepresentation()
self.assertEqual(
newRepresentation,
representation,
"The new pattern did not map to the same object representation"
)
# check that the pattern was correctly learnt and is stable
self.pooler.reset()
self.infer(feedforwardPattern=object[0])
self.assertEqual(
self._getActiveRepresentation(),
representation,
"The pooled representation is not stable"
)
def testLearnSingleObject(self):
"""
Many patterns are learnt for a single object.
Objects: A{P, Q, R, S, T}
"""
self.init()
object = self.generateObject(numPatterns=5)
self.learn(object, numRepetitions=2, randomOrder=True, newObject=True)
representation = self._getActiveRepresentation()
# check that all patterns map to the same object
for pattern in object:
self.pooler.reset()
self.infer(feedforwardPattern=pattern)
self.assertEqual(
self._getActiveRepresentation(),
representation,
"The pooled representation is not stable"
)
# if activity stops, check that the representation persists
self.infer(feedforwardPattern=set())
self.assertEqual(
self._getActiveRepresentation(),
representation,
"The pooled representation did not persist"
)
def testLearnTwoObjectNoCommonPattern(self):
"""
Same test as before, using two objects, without common pattern.
Objects: A{P, Q, R, S,T} B{V, W, X, Y, Z}
"""
self.init()
objectA = self.generateObject(numPatterns=5)
self.learn(objectA, numRepetitions=3, randomOrder=True, newObject=True)
representationA = self._getActiveRepresentation()
objectB = self.generateObject(numPatterns=5)
self.learn(objectB, numRepetitions=3, randomOrder=True, newObject=True)
representationB = self._getActiveRepresentation()
self.assertNotEqual(representationA, representationB)
# check that all patterns map to the same object
for pattern in objectA:
self.pooler.reset()
self.infer(feedforwardPattern=pattern)
self.assertEqual(
self._getActiveRepresentation(),
representationA,
"The pooled representation for the first object is not stable"
)
# check that all patterns map to the same object
for pattern in objectB:
self.pooler.reset()
self.infer(feedforwardPattern=pattern)
self.assertEqual(
self._getActiveRepresentation(),
representationB,
"The pooled representation for the second object is not stable"
)
# feed union of patterns in object A
pattern = objectA[0] | objectA[1]
self.pooler.reset()
self.infer(feedforwardPattern=pattern)
self.assertEqual(
self._getActiveRepresentation(),
representationA,
"The active representation is incorrect"
)
# feed unions of patterns in objects A and B
pattern = objectA[0] | objectB[0]
self.pooler.reset()
self.infer(feedforwardPattern=pattern)
self.assertEqual(
self._getActiveRepresentation(),
representationA | representationB,
"The active representation is incorrect"
)
def testLearnTwoObjectsOneCommonPattern(self):
"""
Same test as before, except the two objects share a pattern
Objects: A{P, Q, R, S,T} B{P, W, X, Y, Z}
"""
self.init()
objectA = self.generateObject(numPatterns=5)
self.learn(objectA, numRepetitions=3, randomOrder=True, newObject=True)
representationA = self._getActiveRepresentation()
objectB = self.generateObject(numPatterns=5)
objectB[0] = objectA[0]
self.learn(objectB, numRepetitions=3, randomOrder=True, newObject=True)
representationB = self._getActiveRepresentation()
self.assertNotEqual(representationA, representationB)
# very small overlap
self.assertLessEqual(len(representationA & representationB), 3)
# check that all patterns except the common one map to the same object
for pattern in objectA[1:]:
self.pooler.reset()
self.infer(feedforwardPattern=pattern)
self.assertEqual(
self._getActiveRepresentation(),
representationA,
"The pooled representation for the first object is not stable"
)
# check that all patterns except the common one map to the same object
for pattern in objectB[1:]:
self.pooler.reset()
self.infer(feedforwardPattern=pattern)
self.assertEqual(
self._getActiveRepresentation(),
representationB,
"The pooled representation for the second object is not stable"
)
# feed shared pattern
pattern = objectA[0]
self.pooler.reset()
self.infer(feedforwardPattern=pattern)
self.assertEqual(
self._getActiveRepresentation(),
representationA | representationB,
"The active representation is incorrect"
)
# feed union of patterns in object A
pattern = objectA[1] | objectA[2]
self.pooler.reset()
self.infer(feedforwardPattern=pattern)
self.assertEqual(
self._getActiveRepresentation(),
representationA,
"The active representation is incorrect"
)
# feed unions of patterns in objects A and B
pattern = objectA[1] | objectB[1]
self.pooler.reset()
self.infer(feedforwardPattern=pattern)
self.assertEqual(
self._getActiveRepresentation(),
representationA | representationB,
"The active representation is incorrect"
)
def testLearnThreeObjectsOneCommonPattern(self):
"""
Same test as before, with three objects
Objects: A{P, Q, R, S,T} B{P, W, X, Y, Z} C{W, H, I, K, L}
"""
self.init()
objectA = self.generateObject(numPatterns=5)
self.learn(objectA, numRepetitions=3, randomOrder=True, newObject=True)
representationA = self._getActiveRepresentation()
objectB = self.generateObject(numPatterns=5)
objectB[0] = objectA[0]
self.learn(objectB, numRepetitions=3, randomOrder=True, newObject=True)
representationB = self._getActiveRepresentation()
objectC = self.generateObject(numPatterns=5)
objectC[0] = objectB[1]
self.learn(objectC, numRepetitions=3, randomOrder=True, newObject=True)
representationC = self._getActiveRepresentation()
self.assertNotEquals(representationA, representationB, representationC)
# very small overlap
self.assertLessEqual(len(representationA & representationB), 3)
self.assertLessEqual(len(representationB & representationC), 3)
self.assertLessEqual(len(representationA & representationC), 3)
# check that all patterns except the common one map to the same object
for pattern in objectA[1:]:
self.pooler.reset()
self.infer(feedforwardPattern=pattern)
self.assertEqual(
self._getActiveRepresentation(),
representationA,
"The pooled representation for the first object is not stable"
)
# check that all patterns except the common one map to the same object
for pattern in objectB[2:]:
self.pooler.reset()
self.infer(feedforwardPattern=pattern)
self.assertEqual(
self._getActiveRepresentation(),
representationB,
"The pooled representation for the second object is not stable"
)
# check that all patterns except the common one map to the same object
for pattern in objectC[1:]:
self.pooler.reset()
self.infer(feedforwardPattern=pattern)
self.assertEqual(
self._getActiveRepresentation(),
representationC,
"The pooled representation for the third object is not stable"
)
# feed shared pattern between A and B
pattern = objectA[0]
self.pooler.reset()
self.infer(feedforwardPattern=pattern)
self.assertEqual(
self._getActiveRepresentation(),
representationA | representationB,
"The active representation is incorrect"
)
# feed shared pattern between B and C
pattern = objectB[1]
self.pooler.reset()
self.infer(feedforwardPattern=pattern)
self.assertEqual(
self._getActiveRepresentation(),
representationB | representationC,
"The active representation is incorrect"
)
# feed union of patterns in object A
pattern = objectA[1] | objectA[2]
self.pooler.reset()
self.infer(feedforwardPattern=pattern)
self.assertEqual(
self._getActiveRepresentation(),
representationA,
"The active representation is incorrect"
)
# feed unions of patterns to activate all objects
pattern = objectA[1] | objectB[1]
self.pooler.reset()
self.infer(feedforwardPattern=pattern)
self.assertEqual(
self._getActiveRepresentation(),
representationA | representationB | representationC,
"The active representation is incorrect"
)
def testLearnThreeObjectsOneCommonPatternSpatialNoise(self):
"""
Same test as before, with three objects
Objects: A{P, Q, R, S,T} B{P, W, X, Y, Z} C{W, H, I, K, L}
"""
self.init()
objectA = self.generateObject(numPatterns=5)
self.learn(objectA, numRepetitions=3, randomOrder=True, newObject=True)
representationA = self._getActiveRepresentation()
objectB = self.generateObject(numPatterns=5)
objectB[0] = objectA[0]
self.learn(objectB, numRepetitions=3, randomOrder=True, newObject=True)
representationB = self._getActiveRepresentation()
objectC = self.generateObject(numPatterns=5)
objectC[0] = objectB[1]
self.learn(objectC, numRepetitions=3, randomOrder=True, newObject=True)
representationC = self._getActiveRepresentation()
self.assertNotEquals(representationA, representationB, representationC)
# very small overlap
self.assertLessEqual(len(representationA & representationB), 3)
self.assertLessEqual(len(representationB & representationC), 3)
self.assertLessEqual(len(representationA & representationC), 3)
# check that all patterns except the common one map to the same object
for pattern in objectA[1:]:
noisyPattern = self.proximalPatternMachine.addNoise(pattern, 0.05)
self.pooler.reset()
self.infer(feedforwardPattern=noisyPattern)
self.assertEqual(
self._getActiveRepresentation(),
representationA,
"The pooled representation for the first object is not stable"
)
# check that all patterns except the common one map to the same object
for pattern in objectB[2:]:
noisyPattern = self.proximalPatternMachine.addNoise(pattern, 0.05)
self.pooler.reset()
self.infer(feedforwardPattern=noisyPattern)
self.assertEqual(
self._getActiveRepresentation(),
representationB,
"The pooled representation for the second object is not stable"
)
# check that all patterns except the common one map to the same object
for pattern in objectC[1:]:
noisyPattern = self.proximalPatternMachine.addNoise(pattern, 0.05)
self.pooler.reset()
self.infer(feedforwardPattern=noisyPattern)
self.assertEqual(
self._getActiveRepresentation(),
representationC,
"The pooled representation for the third object is not stable"
)
# feed shared pattern between A and B
pattern = objectA[0]
noisyPattern = self.proximalPatternMachine.addNoise(pattern, 0.05)
self.pooler.reset()
self.infer(feedforwardPattern=noisyPattern)
self.assertEqual(
self._getActiveRepresentation(),
representationA | representationB,
"The active representation is incorrect"
)
# feed shared pattern between B and C
pattern = objectB[1]
noisyPattern = self.proximalPatternMachine.addNoise(pattern, 0.05)
self.pooler.reset()
self.infer(feedforwardPattern=noisyPattern)
self.assertEqual(
self._getActiveRepresentation(),
representationB | representationC,
"The active representation is incorrect"
)
# feed union of patterns in object A
pattern = objectA[1] | objectA[2]
noisyPattern = self.proximalPatternMachine.addNoise(pattern, 0.05)
self.pooler.reset()
self.infer(feedforwardPattern=noisyPattern)
self.assertEqual(
self._getActiveRepresentation(),
representationA,
"The active representation is incorrect"
)
# feed unions of patterns to activate all objects
pattern = objectA[1] | objectB[1]
noisyPattern = self.proximalPatternMachine.addNoise(pattern, 0.05)
self.pooler.reset()
self.infer(feedforwardPattern=noisyPattern)
self.assertEqual(
self._getActiveRepresentation(),
representationA | representationB | representationC,
"The active representation is incorrect"
)
def testInferObjectOverTime(self):
"""Infer an object after touching only ambiguous points."""
self.init()
patterns = [self.generatePattern() for _ in xrange(3)]
objectA = [patterns[0], patterns[1]]
objectB = [patterns[1], patterns[2]]
objectC = [patterns[2], patterns[0]]
self.learn(objectA, numRepetitions=3, newObject=True)
representationA = set(self.pooler.getActiveCells())
self.learn(objectB, numRepetitions=3, newObject=True)
representationB = set(self.pooler.getActiveCells())
self.learn(objectC, numRepetitions=3, newObject=True)
representationC = set(self.pooler.getActiveCells())
self.pooler.reset()
self.infer(patterns[0])
self.assertEqual(set(self.pooler.getActiveCells()),
representationA | representationC)
self.infer(patterns[1])
self.assertEqual(set(self.pooler.getActiveCells()),
representationA)
def testLearnOneObjectInTwoColumns(self):
"""Learns one object in two different columns."""
self.init(numCols=2)
neighborsIndices = [[1], [0]]
objectA = self.generateObject(numPatterns=5, numCols=2)
# learn object
self.learnMultipleColumns(
objectA,
numRepetitions=3,
neighborsIndices=neighborsIndices,
randomOrder=True,
newObject=True
)
objectARepresentations = self._getActiveRepresentations()
for pooler in self.poolers:
pooler.reset()
for patterns in objectA:
for i in xrange(3):
activeRepresentations = self._getActiveRepresentations()
self.inferMultipleColumns(
feedforwardPatterns=patterns,
activeRepresentations=activeRepresentations,
neighborsIndices=neighborsIndices,
)
if i > 0:
self.assertEqual(activeRepresentations,
self._getActiveRepresentations())
self.assertEqual(objectARepresentations,
self._getActiveRepresentations())
def testLearnTwoObjectsInTwoColumnsNoCommonPattern(self):
"""Learns two objects in two different columns."""
self.init(numCols=2)
neighborsIndices = [[1], [0]]
objectA = self.generateObject(numPatterns=5, numCols=2)
objectB = self.generateObject(numPatterns=5, numCols=2)
# learn object
self.learnMultipleColumns(
objectA,
numRepetitions=3,
neighborsIndices=neighborsIndices,
randomOrder=True,
newObject=True
)
activeRepresentationsA = self._getActiveRepresentations()
# learn object
self.learnMultipleColumns(
objectB,
numRepetitions=3,
neighborsIndices=neighborsIndices,
randomOrder=True,
newObject=True,
)
activeRepresentationsB = self._getActiveRepresentations()
for pooler in self.poolers:
pooler.reset()
# check inference for object A
for patternsA in objectA:
for i in xrange(3):
activeRepresentations = self._getActiveRepresentations()
self.inferMultipleColumns(
feedforwardPatterns=patternsA,
activeRepresentations=activeRepresentations,
neighborsIndices=neighborsIndices,
)
self.assertEqual(
activeRepresentationsA,
self._getActiveRepresentations()
)
for pooler in self.poolers:
pooler.reset()
# check inference for object B
for patternsB in objectB:
for i in xrange(3):
activeRepresentations = self._getActiveRepresentations()
self.inferMultipleColumns(
feedforwardPatterns=patternsB,
activeRepresentations=activeRepresentations,
neighborsIndices=neighborsIndices
)
self.assertEqual(
activeRepresentationsB,
self._getActiveRepresentations()
)
def testLearnTwoObjectsInTwoColumnsOneCommonPattern(self):
"""Learns two objects in two different columns, with a common pattern."""
self.init(numCols=2)
neighborsIndices = [[1], [0]]
objectA = self.generateObject(numPatterns=5, numCols=2)
objectB = self.generateObject(numPatterns=5, numCols=2)
# second pattern in column 0 is shared
objectB[1][0] = objectA[1][0]
# learn object
self.learnMultipleColumns(
objectA,
numRepetitions=3,
neighborsIndices=neighborsIndices,
randomOrder=True,
newObject=True
)
activeRepresentationsA = self._getActiveRepresentations()
# learn object
self.learnMultipleColumns(
objectB,
numRepetitions=3,
neighborsIndices=neighborsIndices,
randomOrder=True,
newObject=True
)
activeRepresentationsB = self._getActiveRepresentations()
# check inference for object A
# for the first pattern, the distal predictions won't be correct
# for the second one, the prediction will be unique thanks to the
# distal predictions from the other column which has no ambiguity
for pooler in self.poolers:
pooler.reset()
for patternsA in objectA:
for i in xrange(3):
activeRepresentations = self._getActiveRepresentations()
self.inferMultipleColumns(
feedforwardPatterns=patternsA,
activeRepresentations=activeRepresentations,
neighborsIndices=neighborsIndices,
)
self.assertEqual(
activeRepresentationsA,
self._getActiveRepresentations()
)
for pooler in self.poolers:
pooler.reset()
# check inference for object B
for patternsB in objectB:
for i in xrange(3):
activeRepresentations = self._getActiveRepresentations()
self.inferMultipleColumns(
feedforwardPatterns=patternsB,
activeRepresentations=activeRepresentations,
neighborsIndices=neighborsIndices
)
self.assertEqual(
activeRepresentationsB,
self._getActiveRepresentations()
)
def testLearnTwoObjectsInTwoColumnsOneCommonPatternEmptyFirstInput(self):
"""Learns two objects in two different columns, with a common pattern."""
self.init(numCols=2)
neighborsIndices = [[1], [0]]
objectA = self.generateObject(numPatterns=5, numCols=2)
objectB = self.generateObject(numPatterns=5, numCols=2)
# second pattern in column 0 is shared
objectB[1][0] = objectA[1][0]
# learn object
self.learnMultipleColumns(
objectA,
numRepetitions=3,
neighborsIndices=neighborsIndices,
randomOrder=True,
newObject=True
)
activeRepresentationsA = self._getActiveRepresentations()
# learn object
self.learnMultipleColumns(
objectB,
numRepetitions=3,
neighborsIndices=neighborsIndices,
randomOrder=True,
newObject=True
)
activeRepresentationsB = self._getActiveRepresentations()
# check inference for object A
for pooler in self.poolers:
pooler.reset()
firstPattern = True
for patternsA in objectA:
activeRepresentations = self._getActiveRepresentations()
if firstPattern:
self.inferMultipleColumns(
feedforwardPatterns=[set(), patternsA[1]],
activeRepresentations=activeRepresentations,
neighborsIndices=neighborsIndices,
)
desiredRepresentation = [set(), activeRepresentationsA[1]]
else:
self.inferMultipleColumns(
feedforwardPatterns=patternsA,
activeRepresentations=activeRepresentations,
neighborsIndices=neighborsIndices,
)
desiredRepresentation = activeRepresentationsA
self.assertEqual(
desiredRepresentation,
self._getActiveRepresentations()
)
def testPersistence(self):
"""After learning, representation should persist in L2 without input."""
self.init(numCols=2)
neighborsIndices = [[1], [0]]
objectA = self.generateObject(numPatterns=5, numCols=2)
# learn object
self.learnMultipleColumns(
objectA,
numRepetitions=3,
neighborsIndices=neighborsIndices,
randomOrder=True,
newObject=True
)
objectARepresentations = self._getActiveRepresentations()
for pooler in self.poolers:
pooler.reset()
for patterns in objectA:
for i in xrange(3):
# replace third pattern for column 2 by empty pattern
if i == 2:
patterns[1] = set()
activeRepresentations = self._getActiveRepresentations()
self.inferMultipleColumns(
feedforwardPatterns=patterns,
activeRepresentations=activeRepresentations,
neighborsIndices=neighborsIndices,
)
if i > 0:
self.assertEqual(activeRepresentations,
self._getActiveRepresentations())
self.assertEqual(objectARepresentations,
self._getActiveRepresentations())
def testLateralDisambiguation(self):
"""Lateral disambiguation using a constant simulated distal input."""
self.init(overrides={
"lateralInputWidths": [self.inputWidth],
})
objectA = self.generateObject(numPatterns=5)
lateralInputA = [[set()]] + [[self.generatePattern()] for _ in xrange(4)]
self.learn(objectA,
lateralPatterns=lateralInputA,
numRepetitions=3,
randomOrder=True,
newObject=True)
representationA = self._getActiveRepresentation()
objectB = self.generateObject(numPatterns=5)
objectB[3] = objectA[3]
lateralInputB = [[set()]] + [[self.generatePattern()] for _ in xrange(4)]
self.learn(objectB,
lateralPatterns=lateralInputB,
numRepetitions=3,
randomOrder=True,
newObject=True)
representationB = self._getActiveRepresentation()
self.assertNotEqual(representationA, representationB)
# very small overlap
self.assertLessEqual(len(representationA & representationB), 3)
# no ambiguity with lateral input
for pattern in objectA:
self.pooler.reset()
self.infer(feedforwardPattern=pattern, lateralInputs=lateralInputA[-1])
self.assertEqual(
self._getActiveRepresentation(),
representationA,
"The pooled representation for the first object is not stable"
)
# no ambiguity with lateral input
for pattern in objectB:
self.pooler.reset()
self.infer(feedforwardPattern=pattern, lateralInputs=lateralInputB[-1])
self.assertEqual(
self._getActiveRepresentation(),
representationB,
"The pooled representation for the second object is not stable"
)
def testLateralContestResolved(self):
"""
Infer an object via lateral disambiguation even if some other columns have
similar ambiguity.
"""
self.init(overrides={"lateralInputWidths": [self.inputWidth,
self.inputWidth]})
patterns = [self.generatePattern() for _ in xrange(3)]
objectA = [patterns[0], patterns[1]]
objectB = [patterns[1], patterns[2]]
lateralInput1A = self.generatePattern()
lateralInput2A = self.generatePattern()
lateralInput1B = self.generatePattern()
lateralInput2B = self.generatePattern()
self.learn(objectA, lateralPatterns=[[lateralInput1A, lateralInput2A]]*2,
numRepetitions=3, newObject=True)
representationA = set(self.pooler.getActiveCells())
self.learn(objectB, lateralPatterns=[[lateralInput1B, lateralInput2B]]*2,
numRepetitions=3, newObject=True)
representationB = set(self.pooler.getActiveCells())
self.pooler.reset()
# This column will say A | B
# One lateral column says A | B
# Another lateral column says A
self.infer(patterns[1], lateralInputs=[(), ()])
self.infer(patterns[1], lateralInputs=[lateralInput1A | lateralInput1B,
lateralInput2A])
self.assertEqual(set(self.pooler.getActiveCells()), representationA)
@unittest.skip("Fails, need to discuss")
def testMultiColumnCompetition(self):
"""Competition between multiple conflicting lateral inputs."""
self.init(numCols=4)
neighborsIndices = [[1, 2, 3], [0, 2, 3], [0, 1, 3], [0, 1, 2]]
objectA = self.generateObject(numPatterns=5, numCols=4)
objectB = self.generateObject(numPatterns=5, numCols=4)
# second pattern in column 0 is shared
objectB[1][0] = objectA[1][0]
# learn object
self.learnMultipleColumns(
objectA,
numRepetitions=3,
neighborsIndices=neighborsIndices,
randomOrder=True,
newObject=True
)
activeRepresentationsA = self._getActiveRepresentations()
# learn object
self.learnMultipleColumns(
objectB,
numRepetitions=3,
neighborsIndices=neighborsIndices,
randomOrder=True,
newObject=True
)
activeRepresentationsB = self._getActiveRepresentations()
# check inference for object A
# for the first pattern, the distal predictions won't be correct
# for the second one, the prediction will be unique thanks to the
# distal predictions from the other column which has no ambiguity
for pooler in self.poolers:
pooler.reset()
# sensed patterns will be mixed
sensedPatterns = objectA[1][:-1] + [objectA[1][-1] | objectB[1][-1]]
# feed sensed patterns first time
# every one feels the correct object, except first column which feels
# the union (reminder: lateral input are delayed)
activeRepresentations = self._getActiveRepresentations()
self.inferMultipleColumns(
feedforwardPatterns=sensedPatterns,
activeRepresentations=activeRepresentations,
neighborsIndices=neighborsIndices,
)
firstSensedRepresentations = [
activeRepresentationsA[0] | activeRepresentationsB[0],
activeRepresentationsA[1],
activeRepresentationsA[2],
activeRepresentationsA[3] | activeRepresentationsB[3]
]
self.assertEqual(
firstSensedRepresentations,
self._getActiveRepresentations()
)
# feed sensed patterns second time
# the distal predictions are still ambiguous in C1, but disambiguated
# in C4
activeRepresentations = self._getActiveRepresentations()
self.inferMultipleColumns(
feedforwardPatterns=sensedPatterns,
activeRepresentations=activeRepresentations,
neighborsIndices=neighborsIndices,
)
secondSensedRepresentations = [
activeRepresentationsA[0] | activeRepresentationsB[0],
activeRepresentationsA[1],
activeRepresentationsA[2],
activeRepresentationsA[3]
]
self.assertEqual(
secondSensedRepresentations,
self._getActiveRepresentations()
)
# feed sensed patterns third time
# this time, it is all disambiguated
activeRepresentations = self._getActiveRepresentations()
self.inferMultipleColumns(
feedforwardPatterns=sensedPatterns,
activeRepresentations=activeRepresentations,
neighborsIndices=neighborsIndices,
)
self.assertEqual(
activeRepresentationsA,
self._getActiveRepresentations()
)
def testMutualDisambiguationThroughUnions(self):
"""
Learns three object in two different columns.
Feed ambiguous sensations, A u B and B u C. The system should narrow down
to B.
"""
self.init(numCols=2)
neighborsIndices = [[1], [0]]
objectA = self.generateObject(numPatterns=5, numCols=2)
objectB = self.generateObject(numPatterns=5, numCols=2)
objectC = self.generateObject(numPatterns=5, numCols=2)
# learn object
self.learnMultipleColumns(
objectA,
numRepetitions=3,
neighborsIndices=neighborsIndices,
randomOrder=True,
newObject=True
)
activeRepresentationsA = self._getActiveRepresentations()
# learn object
self.learnMultipleColumns(
objectB,
numRepetitions=3,
neighborsIndices=neighborsIndices,
randomOrder=True,
newObject=True
)
activeRepresentationsB = self._getActiveRepresentations()
# learn object
self.learnMultipleColumns(
objectC,
numRepetitions=3,
neighborsIndices=neighborsIndices,
randomOrder=True,
newObject=True
)
activeRepresentationsC = self._getActiveRepresentations()
# create sensed patterns (ambiguous)
sensedPatterns = [objectA[1][0] | objectB[1][0],
objectB[2][1] | objectC[2][1]]
for pooler in self.poolers:
pooler.reset()
# feed sensed patterns first time
# the L2 representations should be ambiguous
activeRepresentations = self._getActiveRepresentations()
self.inferMultipleColumns(
feedforwardPatterns=sensedPatterns,
activeRepresentations=activeRepresentations,
neighborsIndices=neighborsIndices,
)
firstRepresentations = [
activeRepresentationsA[0] | activeRepresentationsB[0],
activeRepresentationsB[1] | activeRepresentationsC[1]
]
self.assertEqual(
firstRepresentations,
self._getActiveRepresentations()
)
# feed a second time, distal predictions should disambiguate
activeRepresentations = self._getActiveRepresentations()
self.inferMultipleColumns(
feedforwardPatterns=sensedPatterns,
activeRepresentations=activeRepresentations,
neighborsIndices=neighborsIndices,
)
# check that representations are unique, being slightly tolerant
self.assertLessEqual(
len(self._getActiveRepresentations()[0] - activeRepresentationsB[0]),
5,
)
self.assertLessEqual(
len(self._getActiveRepresentations()[1] - activeRepresentationsB[1]),
5,
)
self.assertGreaterEqual(
len(self._getActiveRepresentations()[0] & activeRepresentationsB[0]),
35,
)
self.assertGreaterEqual(
len(self._getActiveRepresentations()[1] & activeRepresentationsB[1]),
35,
)
def setUp(self):
"""
Sets up the test.
"""
# single column case
self.pooler = None
# multi column case
self.poolers = []
# create pattern machine
self.proximalPatternMachine = PatternMachine(
n=self.inputWidth,
w=self.numOutputActiveBits,
num=200,
seed=self.seed
)
self.patternId = 0
np.random.seed(self.seed)
# Wrappers around ColumnPooler API
def learn(self,
feedforwardPatterns,
lateralPatterns=None,
numRepetitions=1,
randomOrder=True,
newObject=True):
"""
Parameters:
----------------------------
Learns a single object, with the provided patterns.
@param feedforwardPatterns (list(set))
List of proximal input patterns
@param lateralPatterns (list(list(iterable)))
List of distal input patterns, or None. If no lateral input is
used. The outer list is expected to have the same length as
feedforwardPatterns, whereas each inner list's length is the
number of cortical columns which are distally connected to the
pooler.
@param numRepetitions (int)
Number of times the patterns will be fed
@param randomOrder (bool)
If true, the order of patterns will be shuffled at each
repetition
"""
if newObject:
self.pooler.mmClearHistory()
self.pooler.reset()
# set-up
indices = range(len(feedforwardPatterns))
if lateralPatterns is None:
lateralPatterns = [[] for _ in xrange(len(feedforwardPatterns))]
for _ in xrange(numRepetitions):
if randomOrder:
np.random.shuffle(indices)
for idx in indices:
self.pooler.compute(sorted(feedforwardPatterns[idx]),
[sorted(lateralPattern)
for lateralPattern in lateralPatterns[idx]],
learn=True)
def infer(self,
feedforwardPattern,
lateralInputs=(),
printMetrics=False):
"""
Feeds a single pattern to the column pooler (as well as an eventual lateral
pattern).
Parameters:
----------------------------
@param feedforwardPattern (set)
Input proximal pattern to the pooler
@param lateralInputs (list(set))
Input distal patterns to the pooler (one for each neighboring CC's)
@param printMetrics (bool)
If true, will print cell metrics
"""
self.pooler.compute(sorted(feedforwardPattern),
[sorted(lateralInput)
for lateralInput in lateralInputs],
learn=False)
if printMetrics:
print self.pooler.mmPrettyPrintMetrics(
self.pooler.mmGetDefaultMetrics()
)
# Helper functions
def generatePattern(self):
"""
Returns a random proximal input pattern.
"""
pattern = self.proximalPatternMachine.get(self.patternId)
self.patternId += 1
return pattern
def generateObject(self, numPatterns, numCols=1):
"""
Creates a list of patterns, for a given object.
If numCols > 1 is given, a list of list of patterns will be returned.
"""
if numCols == 1:
return [self.generatePattern() for _ in xrange(numPatterns)]
else:
patterns = []
for i in xrange(numPatterns):
patterns.append([self.generatePattern() for _ in xrange(numCols)])
return patterns
def init(self, overrides=None, numCols=1):
"""
Creates the column pooler with specified parameter overrides.
Except for the specified overrides and problem-specific parameters, used
parameters are implementation defaults.
"""
params = {
"inputWidth": self.inputWidth,
"lateralInputWidths": [self.outputWidth]*(numCols-1),
"cellCount": self.outputWidth,
"sdrSize": self.numOutputActiveBits,
"minThresholdProximal": 10,
"sampleSizeProximal": 20,
"connectedPermanenceProximal": 0.6,
"initialDistalPermanence": 0.51,
"activationThresholdDistal": 10,
"sampleSizeDistal": 20,
"connectedPermanenceDistal": 0.6,
"seed": self.seed,
}
if overrides is None:
overrides = {}
params.update(overrides)
if numCols == 1:
self.pooler = MonitoredColumnPooler(**params)
else:
# TODO: We need a different seed for each pooler otherwise each one
# outputs an identical representation. Use random seed for now but ideally
# we would set different specific seeds for each pooler
params['seed']=0
self.poolers = [MonitoredColumnPooler(**params) for _ in xrange(numCols)]
def _getActiveRepresentation(self):
"""
Retrieves the current active representation in the pooler.
"""
if self.pooler is None:
raise ValueError("No pooler has been instantiated")
return set(self.pooler.getActiveCells())
# Multi-column testing
def learnMultipleColumns(self,
feedforwardPatterns,
numRepetitions=1,
neighborsIndices=None,
randomOrder=True,
newObject=True):
"""
Learns a single object, feeding it through the multiple columns.
Parameters:
----------------------------
Learns a single object, with the provided patterns.
@param feedforwardPatterns (list(list(set)))
List of proximal input patterns (one for each pooler).
@param neighborsIndices (list(list))
List of column indices each column received input from.
@param numRepetitions (int)
Number of times the patterns will be fed
@param randomOrder (bool)
If true, the order of patterns will be shuffled at each
repetition
"""
if newObject:
for pooler in self.poolers:
pooler.mmClearHistory()
pooler.reset()
# use different set of pattern indices to allow random orders
indices = [range(len(feedforwardPatterns))] * len(self.poolers)
prevActiveCells = [set() for _ in xrange(len(self.poolers))]
# by default, all columns are neighbors
if neighborsIndices is None:
neighborsIndices = [
range(i) + range(i+1, len(self.poolers))
for i in xrange(len(self.poolers))
]
for _ in xrange(numRepetitions):
# independently shuffle pattern orders if necessary
if randomOrder:
for idx in indices:
np.random.shuffle(idx)
for i in xrange(len(indices[0])):
# Train each column
for col, pooler in enumerate(self.poolers):
# get union of relevant lateral representations
lateralInputs = [sorted(activeCells)
for presynapticCol, activeCells
in enumerate(prevActiveCells)
if col != presynapticCol]
pooler.compute(sorted(feedforwardPatterns[indices[col][i]][col]),
lateralInputs, learn=True)
prevActiveCells = self._getActiveRepresentations()
def inferMultipleColumns(self,
feedforwardPatterns,
activeRepresentations,
neighborsIndices=None,
printMetrics=False,
reset=False):
"""
Feeds a single pattern to the column pooler (as well as an eventual lateral
pattern).
Parameters:
----------------------------
@param feedforwardPattern (list(set))
Input proximal patterns to the pooler (one for each column)
@param activeRepresentations (list(set))
Active representations in the columns at the previous step.
@param neighborsIndices (list(list))
List of column indices each column received input from.
@param printMetrics (bool)
If true, will print cell metrics
"""
if reset:
for pooler in self.poolers:
pooler.reset()
# by default, all columns are neighbors
if neighborsIndices is None:
neighborsIndices = [
range(i) + range(i+1, len(self.poolers))
for i in xrange(len(self.poolers))
]
for col, pooler in enumerate(self.poolers):
# get union of relevant lateral representations
lateralInputs = [sorted(activeCells)
for presynapticCol, activeCells
in enumerate(activeRepresentations)
if col != presynapticCol]
pooler.compute(sorted(feedforwardPatterns[col]),
lateralInputs, learn=False)
if printMetrics:
for pooler in self.poolers:
print pooler.mmPrettyPrintMetrics(
pooler.mmGetDefaultMetrics()
)
def _getActiveRepresentations(self):
"""
Retrieves the current active representations in the poolers.
"""
if len(self.poolers) == 0:
raise ValueError("No pooler has been instantiated")
return [set(pooler.getActiveCells()) for pooler in self.poolers]
def setUp(self): self.patternMachine = PatternMachine(10000, 5, num=50)
class PatternMachineTest(unittest.TestCase):
def setUp(self):
self.patternMachine = PatternMachine(10000, 5, num=50)
def testGet(self):
patternA = self.patternMachine.get(48)
self.assertEqual(len(patternA), 5)
patternB = self.patternMachine.get(49)
self.assertEqual(len(patternB), 5)
self.assertEqual(patternA & patternB, set())
def testGetOutOfBounds(self):
args = [50]
self.assertRaises(IndexError, self.patternMachine.get, *args)
def testAddNoise(self):
patternMachine = PatternMachine(10000, 1000, num=1)
pattern = patternMachine.get(0)
noisy = patternMachine.addNoise(pattern, 0.0)
self.assertEqual(len(pattern & noisy), 1000)
noisy = patternMachine.addNoise(pattern, 0.5)
self.assertTrue(400 < len(pattern & noisy) < 600)
noisy = patternMachine.addNoise(pattern, 1.0)
self.assertTrue(50 < len(pattern & noisy) < 150)
def testNumbersForBit(self):
pattern = self.patternMachine.get(49)
for bit in pattern:
self.assertEqual(self.patternMachine.numbersForBit(bit), set([49]))
def testNumbersForBitOutOfBounds(self):
args = [10000]
self.assertRaises(IndexError, self.patternMachine.numbersForBit, *args)
def testNumberMapForBits(self):
pattern = self.patternMachine.get(49)
numberMap = self.patternMachine.numberMapForBits(pattern)
self.assertEqual(numberMap.keys(), [49])
self.assertEqual(numberMap[49], pattern)
def testWList(self):
w = [4, 7, 11]
patternMachine = PatternMachine(100, w, num=50)
widths = dict((el, 0) for el in w)
for i in range(50):
pattern = patternMachine.get(i)
width = len(pattern)
self.assertTrue(width in w)
widths[len(pattern)] += 1
for i in w:
self.assertTrue(widths[i] > 0)
class TemporalMemoryTest(unittest.TestCase):
def setUp(self):
self.tm = TemporalMemory()
def testInitInvalidParams(self):
# Invalid columnDimensions
kwargs = {"columnDimensions": [], "cellsPerColumn": 32}
self.assertRaises(ValueError, TemporalMemory, **kwargs)
# Invalid cellsPerColumn
kwargs = {"columnDimensions": [2048], "cellsPerColumn": 0}
self.assertRaises(ValueError, TemporalMemory, **kwargs)
kwargs = {"columnDimensions": [2048], "cellsPerColumn": -10}
self.assertRaises(ValueError, TemporalMemory, **kwargs)
def testActivateCorrectlyPredictiveCells(self):
tm = self.tm
prevPredictiveCells = set([0, 237, 1026, 26337, 26339, 55536])
activeColumns = set([32, 47, 823])
prevMatchingCells = set()
(activeCells, winnerCells, predictedColumns,
predictedInactiveCells) = tm.activateCorrectlyPredictiveCells(
prevPredictiveCells, prevMatchingCells, activeColumns)
self.assertEqual(activeCells, set([1026, 26337, 26339]))
self.assertEqual(winnerCells, set([1026, 26337, 26339]))
self.assertEqual(predictedColumns, set([32, 823]))
self.assertEqual(predictedInactiveCells, set())
def testActivateCorrectlyPredictiveCellsEmpty(self):
tm = self.tm
# No previous predictive cells, no active columns
prevPredictiveCells = set()
activeColumns = set()
prevMatchingCells = set()
(activeCells, winnerCells, predictedColumns,
predictedInactiveCells) = tm.activateCorrectlyPredictiveCells(
prevPredictiveCells, prevMatchingCells, activeColumns)
self.assertEqual(activeCells, set())
self.assertEqual(winnerCells, set())
self.assertEqual(predictedColumns, set())
self.assertEqual(predictedInactiveCells, set())
# No previous predictive cells, with active columns
prevPredictiveCells = set()
activeColumns = set([32, 47, 823])
prevMatchingCells = set()
(activeCells, winnerCells, predictedColumns,
predictedInactiveCells) = tm.activateCorrectlyPredictiveCells(
prevPredictiveCells, prevMatchingCells, activeColumns)
self.assertEqual(activeCells, set())
self.assertEqual(winnerCells, set())
self.assertEqual(predictedColumns, set())
self.assertEqual(predictedInactiveCells, set())
# No active columns, with previously predictive cells
prevPredictiveCells = set([0, 237, 1026, 26337, 26339, 55536])
activeColumns = set()
prevMatchingCells = set()
(activeCells, winnerCells, predictedColumns,
predictedInactiveCells) = tm.activateCorrectlyPredictiveCells(
prevPredictiveCells, prevMatchingCells, activeColumns)
self.assertEqual(activeCells, set())
self.assertEqual(winnerCells, set())
self.assertEqual(predictedColumns, set())
self.assertEqual(predictedInactiveCells, set())
def testActivateCorrectlyPredictiveCellsOrphan(self):
tm = self.tm
tm.predictedSegmentDecrement = 0.001
prevPredictiveCells = set([])
activeColumns = set([32, 47, 823])
prevMatchingCells = set([32, 47])
(activeCells, winnerCells, predictedColumns,
predictedInactiveCells) = tm.activateCorrectlyPredictiveCells(
prevPredictiveCells, prevMatchingCells, activeColumns)
self.assertEqual(activeCells, set([]))
self.assertEqual(winnerCells, set([]))
self.assertEqual(predictedColumns, set([]))
self.assertEqual(predictedInactiveCells, set([32, 47]))
def testBurstColumns(self):
tm = TemporalMemory(cellsPerColumn=4,
connectedPermanence=0.50,
minThreshold=1,
seed=42)
connections = tm.connections
connections.createSegment(0)
connections.createSynapse(0, 23, 0.6)
connections.createSynapse(0, 37, 0.4)
connections.createSynapse(0, 477, 0.9)
connections.createSegment(0)
connections.createSynapse(1, 49, 0.9)
connections.createSynapse(1, 3, 0.8)
connections.createSegment(1)
connections.createSynapse(2, 733, 0.7)
connections.createSegment(108)
connections.createSynapse(3, 486, 0.9)
activeColumns = set([0, 1, 26])
predictedColumns = set([26])
prevActiveCells = set([23, 37, 49, 733])
prevWinnerCells = set([23, 37, 49, 733])
(activeCells, winnerCells,
learningSegments) = tm.burstColumns(activeColumns, predictedColumns,
prevActiveCells, prevWinnerCells,
connections)
self.assertEqual(activeCells, set([0, 1, 2, 3, 4, 5, 6, 7]))
self.assertEqual(winnerCells, set([0, 6])) # 6 is randomly chosen cell
self.assertEqual(learningSegments,
set([0, 4])) # 4 is new segment created
# Check that new segment was added to winner cell (6) in column 1
self.assertEqual(connections.segmentsForCell(6), set([4]))
def testBurstColumnsEmpty(self):
tm = self.tm
activeColumns = set()
predictedColumns = set()
prevActiveCells = set()
prevWinnerCells = set()
connections = tm.connections
(activeCells, winnerCells,
learningSegments) = tm.burstColumns(activeColumns, predictedColumns,
prevActiveCells, prevWinnerCells,
connections)
self.assertEqual(activeCells, set())
self.assertEqual(winnerCells, set())
self.assertEqual(learningSegments, set())
def testLearnOnSegments(self):
tm = TemporalMemory(maxNewSynapseCount=2)
connections = tm.connections
connections.createSegment(0)
connections.createSynapse(0, 23, 0.6)
connections.createSynapse(0, 37, 0.4)
connections.createSynapse(0, 477, 0.9)
connections.createSegment(1)
connections.createSynapse(1, 733, 0.7)
connections.createSegment(8)
connections.createSynapse(2, 486, 0.9)
connections.createSegment(100)
prevActiveSegments = set([0, 2])
learningSegments = set([1, 3])
prevActiveCells = set([23, 37, 733])
winnerCells = set([0])
prevWinnerCells = set([10, 11, 12, 13, 14])
predictedInactiveCells = set()
prevMatchingSegments = set()
tm.learnOnSegments(prevActiveSegments, learningSegments,
prevActiveCells, winnerCells, prevWinnerCells,
connections, predictedInactiveCells,
prevMatchingSegments)
# Check segment 0
synapseData = connections.dataForSynapse(0)
self.assertAlmostEqual(synapseData.permanence, 0.7)
synapseData = connections.dataForSynapse(1)
self.assertAlmostEqual(synapseData.permanence, 0.5)
synapseData = connections.dataForSynapse(2)
self.assertAlmostEqual(synapseData.permanence, 0.8)
# Check segment 1
synapseData = connections.dataForSynapse(3)
self.assertAlmostEqual(synapseData.permanence, 0.8)
self.assertEqual(len(connections.synapsesForSegment(1)), 2)
# Check segment 2
synapseData = connections.dataForSynapse(4)
self.assertAlmostEqual(synapseData.permanence, 0.9)
self.assertEqual(len(connections.synapsesForSegment(2)), 1)
# Check segment 3
self.assertEqual(len(connections.synapsesForSegment(3)), 2)
def testComputePredictiveCells(self):
tm = TemporalMemory(activationThreshold=2,
minThreshold=2,
predictedSegmentDecrement=0.004)
connections = tm.connections
connections.createSegment(0)
connections.createSynapse(0, 23, 0.6)
connections.createSynapse(0, 37, 0.5)
connections.createSynapse(0, 477, 0.9)
connections.createSegment(1)
connections.createSynapse(1, 733, 0.7)
connections.createSynapse(1, 733, 0.4)
connections.createSegment(1)
connections.createSynapse(2, 974, 0.9)
connections.createSegment(8)
connections.createSynapse(3, 486, 0.9)
connections.createSegment(100)
activeCells = set([23, 37, 733, 974])
(activeSegments, predictiveCells, matchingSegments,
matchingCells) = tm.computePredictiveCells(activeCells, connections)
self.assertEqual(activeSegments, set([0]))
self.assertEqual(predictiveCells, set([0]))
self.assertEqual(matchingSegments, set([0, 1]))
self.assertEqual(matchingCells, set([0, 1]))
def testBestMatchingCell(self):
tm = TemporalMemory(connectedPermanence=0.50, minThreshold=1, seed=42)
connections = tm.connections
connections.createSegment(0)
connections.createSynapse(0, 23, 0.6)
connections.createSynapse(0, 37, 0.4)
connections.createSynapse(0, 477, 0.9)
connections.createSegment(0)
connections.createSynapse(1, 49, 0.9)
connections.createSynapse(1, 3, 0.8)
connections.createSegment(1)
connections.createSynapse(2, 733, 0.7)
connections.createSegment(108)
connections.createSynapse(3, 486, 0.9)
activeCells = set([23, 37, 49, 733])
self.assertEqual(
tm.bestMatchingCell(tm.cellsForColumn(0), activeCells,
connections), (0, 0))
self.assertEqual(
tm.bestMatchingCell(
tm.cellsForColumn(3), # column containing cell 108
activeCells,
connections),
(96, None)) # Random cell from column
self.assertEqual(
tm.bestMatchingCell(tm.cellsForColumn(999), activeCells,
connections),
(31972, None)) # Random cell from column
def testBestMatchingCellFewestSegments(self):
tm = TemporalMemory(columnDimensions=[2],
cellsPerColumn=2,
connectedPermanence=0.50,
minThreshold=1,
seed=42)
connections = tm.connections
connections.createSegment(0)
connections.createSynapse(0, 3, 0.3)
activeSynapsesForSegment = set([])
for _ in range(100):
# Never pick cell 0, always pick cell 1
(cell, _) = tm.bestMatchingCell(tm.cellsForColumn(0),
activeSynapsesForSegment,
connections)
self.assertEqual(cell, 1)
def testBestMatchingSegment(self):
tm = TemporalMemory(connectedPermanence=0.50, minThreshold=1)
connections = tm.connections
connections.createSegment(0)
connections.createSynapse(0, 23, 0.6)
connections.createSynapse(0, 37, 0.4)
connections.createSynapse(0, 477, 0.9)
connections.createSegment(0)
connections.createSynapse(1, 49, 0.9)
connections.createSynapse(1, 3, 0.8)
connections.createSegment(1)
connections.createSynapse(2, 733, 0.7)
connections.createSegment(8)
connections.createSynapse(3, 486, 0.9)
activeCells = set([23, 37, 49, 733])
self.assertEqual(tm.bestMatchingSegment(0, activeCells, connections),
(0, 2))
self.assertEqual(tm.bestMatchingSegment(1, activeCells, connections),
(2, 1))
self.assertEqual(tm.bestMatchingSegment(8, activeCells, connections),
(None, None))
self.assertEqual(tm.bestMatchingSegment(100, activeCells, connections),
(None, None))
def testLeastUsedCell(self):
tm = TemporalMemory(columnDimensions=[2], cellsPerColumn=2, seed=42)
connections = tm.connections
connections.createSegment(0)
connections.createSynapse(0, 3, 0.3)
for _ in range(100):
# Never pick cell 0, always pick cell 1
self.assertEqual(
tm.leastUsedCell(tm.cellsForColumn(0), connections), 1)
def testAdaptSegment(self):
tm = self.tm
connections = tm.connections
connections.createSegment(0)
connections.createSynapse(0, 23, 0.6)
connections.createSynapse(0, 37, 0.4)
connections.createSynapse(0, 477, 0.9)
tm.adaptSegment(0, set([0, 1]), connections, tm.permanenceIncrement,
tm.permanenceDecrement)
synapseData = connections.dataForSynapse(0)
self.assertAlmostEqual(synapseData.permanence, 0.7)
synapseData = connections.dataForSynapse(1)
self.assertAlmostEqual(synapseData.permanence, 0.5)
synapseData = connections.dataForSynapse(2)
self.assertAlmostEqual(synapseData.permanence, 0.8)
def testAdaptSegmentToMax(self):
tm = self.tm
connections = tm.connections
connections.createSegment(0)
connections.createSynapse(0, 23, 0.9)
tm.adaptSegment(0, set([0]), connections, tm.permanenceIncrement,
tm.permanenceDecrement)
synapseData = connections.dataForSynapse(0)
self.assertAlmostEqual(synapseData.permanence, 1.0)
# Now permanence should be at max
tm.adaptSegment(0, set([0]), connections, tm.permanenceIncrement,
tm.permanenceDecrement)
synapseData = connections.dataForSynapse(0)
self.assertAlmostEqual(synapseData.permanence, 1.0)
def testAdaptSegmentToMin(self):
tm = self.tm
connections = tm.connections
connections.createSegment(0)
connections.createSynapse(0, 23, 0.1)
tm.adaptSegment(0, set(), connections, tm.permanenceIncrement,
tm.permanenceDecrement)
synapses = connections.synapsesForSegment(0)
self.assertFalse(0 in synapses)
def testPickCellsToLearnOn(self):
tm = TemporalMemory(seed=42)
connections = tm.connections
connections.createSegment(0)
winnerCells = set([4, 47, 58, 93])
self.assertEqual(tm.pickCellsToLearnOn(2, 0, winnerCells, connections),
set([4, 58])) # randomly picked
self.assertEqual(
tm.pickCellsToLearnOn(100, 0, winnerCells, connections),
set([4, 47, 58, 93]))
self.assertEqual(tm.pickCellsToLearnOn(0, 0, winnerCells, connections),
set())
def testPickCellsToLearnOnAvoidDuplicates(self):
tm = TemporalMemory(seed=42)
connections = tm.connections
connections.createSegment(0)
connections.createSynapse(0, 23, 0.6)
winnerCells = set([23])
# Ensure that no additional (duplicate) cells were picked
self.assertEqual(tm.pickCellsToLearnOn(2, 0, winnerCells, connections),
set())
def testColumnForCell1D(self):
tm = TemporalMemory(columnDimensions=[2048], cellsPerColumn=5)
self.assertEqual(tm.columnForCell(0), 0)
self.assertEqual(tm.columnForCell(4), 0)
self.assertEqual(tm.columnForCell(5), 1)
self.assertEqual(tm.columnForCell(10239), 2047)
def testColumnForCell2D(self):
tm = TemporalMemory(columnDimensions=[64, 64], cellsPerColumn=4)
self.assertEqual(tm.columnForCell(0), 0)
self.assertEqual(tm.columnForCell(3), 0)
self.assertEqual(tm.columnForCell(4), 1)
self.assertEqual(tm.columnForCell(16383), 4095)
def testColumnForCellInvalidCell(self):
tm = TemporalMemory(columnDimensions=[64, 64], cellsPerColumn=4)
try:
tm.columnForCell(16383)
except IndexError:
self.fail("IndexError raised unexpectedly")
args = [16384]
self.assertRaises(IndexError, tm.columnForCell, *args)
args = [-1]
self.assertRaises(IndexError, tm.columnForCell, *args)
def testCellsForColumn1D(self):
tm = TemporalMemory(columnDimensions=[2048], cellsPerColumn=5)
expectedCells = set([5, 6, 7, 8, 9])
self.assertEqual(tm.cellsForColumn(1), expectedCells)
def testCellsForColumn2D(self):
tm = TemporalMemory(columnDimensions=[64, 64], cellsPerColumn=4)
expectedCells = set([256, 257, 258, 259])
self.assertEqual(tm.cellsForColumn(64), expectedCells)
def testCellsForColumnInvalidColumn(self):
tm = TemporalMemory(columnDimensions=[64, 64], cellsPerColumn=4)
try:
tm.cellsForColumn(4095)
except IndexError:
self.fail("IndexError raised unexpectedly")
args = [4096]
self.assertRaises(IndexError, tm.cellsForColumn, *args)
args = [-1]
self.assertRaises(IndexError, tm.cellsForColumn, *args)
def testNumberOfColumns(self):
tm = TemporalMemory(columnDimensions=[64, 64], cellsPerColumn=32)
self.assertEqual(tm.numberOfColumns(), 64 * 64)
def testNumberOfCells(self):
tm = TemporalMemory(columnDimensions=[64, 64], cellsPerColumn=32)
self.assertEqual(tm.numberOfCells(), 64 * 64 * 32)
def testMapCellsToColumns(self):
tm = TemporalMemory(columnDimensions=[100], cellsPerColumn=4)
columnsForCells = tm.mapCellsToColumns(set([0, 1, 2, 5, 399]))
self.assertEqual(columnsForCells[0], set([0, 1, 2]))
self.assertEqual(columnsForCells[1], set([5]))
self.assertEqual(columnsForCells[99], set([399]))
def testWrite(self):
tm1 = TemporalMemory(columnDimensions=[100],
cellsPerColumn=4,
activationThreshold=7,
initialPermanence=0.37,
connectedPermanence=0.58,
minThreshold=4,
maxNewSynapseCount=18,
permanenceIncrement=0.23,
permanenceDecrement=0.08,
seed=91)
# Run some data through before serializing
self.patternMachine = PatternMachine(100, 4)
self.sequenceMachine = SequenceMachine(self.patternMachine)
sequence = self.sequenceMachine.generateFromNumbers(range(5))
for _ in range(3):
for pattern in sequence:
tm1.compute(pattern)
proto1 = TemporalMemoryProto_capnp.TemporalMemoryProto.new_message()
tm1.write(proto1)
# Write the proto to a temp file and read it back into a new proto
with tempfile.TemporaryFile() as f:
proto1.write(f)
f.seek(0)
proto2 = TemporalMemoryProto_capnp.TemporalMemoryProto.read(f)
# Load the deserialized proto
tm2 = TemporalMemory.read(proto2)
# Check that the two temporal memory objects have the same attributes
self.assertEqual(tm1, tm2)
# Run a couple records through after deserializing and check results match
tm1.compute(self.patternMachine.get(0))
tm2.compute(self.patternMachine.get(0))
self.assertEqual(tm1.activeCells, tm2.activeCells)
self.assertEqual(tm1.predictiveCells, tm2.predictiveCells)
self.assertEqual(tm1.winnerCells, tm2.winnerCells)
self.assertEqual(tm1.connections, tm2.connections)
tm1.compute(self.patternMachine.get(3))
tm2.compute(self.patternMachine.get(3))
self.assertEqual(tm1.activeCells, tm2.activeCells)
self.assertEqual(tm1.predictiveCells, tm2.predictiveCells)
self.assertEqual(tm1.winnerCells, tm2.winnerCells)
self.assertEqual(tm1.connections, tm2.connections)
def run(params, paramDir, outputDir, plotVerbosity=0, consoleVerbosity=0):
"""
Runs the union overlap experiment.
:param params: A dict of experiment parameters
:param paramDir: Path of parameter file
:param outputDir: Output will be written to this path
:param plotVerbosity: Plotting verbosity
:param consoleVerbosity: Console output verbosity
"""
print "Running SDR overlap experiment...\n"
print "Params dir: {0}".format(paramDir)
print "Output dir: {0}\n".format(outputDir)
# Dimensionality of sequence patterns
patternDimensionality = params["patternDimensionality"]
# Cardinality (ON / true bits) of sequence patterns
patternCardinality = params["patternCardinality"]
# TODO If this parameter is to be supported, the sequence generation code
# below must change
# Number of unique patterns from which sequences are built
# patternAlphabetSize = params["patternAlphabetSize"]
# Length of sequences shown to network
sequenceLength = params["sequenceLength"]
# Number of sequences used. Sequences may share common elements.
numberOfSequences = params["numberOfSequences"]
# Number of sequence passes for training the TM. Zero => no training.
trainingPasses = params["trainingPasses"]
tmParamOverrides = params["temporalMemoryParams"]
upParamOverrides = params["unionPoolerParams"]
# Generate a sequence list and an associated labeled list (both containing a
# set of sequences separated by None)
start = time.time()
print "\nGenerating sequences..."
patternAlphabetSize = sequenceLength * numberOfSequences
patternMachine = PatternMachine(patternDimensionality, patternCardinality,
patternAlphabetSize)
sequenceMachine = SequenceMachine(patternMachine)
numbers = sequenceMachine.generateNumbers(numberOfSequences, sequenceLength)
generatedSequences = sequenceMachine.generateFromNumbers(numbers)
sequenceLabels = [str(numbers[i + i*sequenceLength: i + (i+1)*sequenceLength])
for i in xrange(numberOfSequences)]
labeledSequences = []
for label in sequenceLabels:
for _ in xrange(sequenceLength):
labeledSequences.append(label)
labeledSequences.append(None)
# Set up the Temporal Memory and Union Pooler network
print "\nCreating network..."
experiment = UnionTemporalPoolerExperiment(tmParamOverrides, upParamOverrides)
# Train only the Temporal Memory on the generated sequences
if trainingPasses > 0:
print "\nTraining Temporal Memory..."
if consoleVerbosity > 0:
print "\nPass\tBursting Columns Mean\tStdDev\tMax"
for i in xrange(trainingPasses):
experiment.runNetworkOnSequences(generatedSequences,
labeledSequences,
tmLearn=True,
upLearn=None,
verbosity=consoleVerbosity,
progressInterval=_SHOW_PROGRESS_INTERVAL)
if consoleVerbosity > 0:
stats = experiment.getBurstingColumnsStats()
print "{0}\t{1}\t{2}\t{3}".format(i, stats[0], stats[1], stats[2])
# Reset the TM monitor mixin's records accrued during this training pass
experiment.tm.mmClearHistory()
print
print MonitorMixinBase.mmPrettyPrintMetrics(
experiment.tm.mmGetDefaultMetrics())
print
if plotVerbosity >= 2:
plotNetworkState(experiment, plotVerbosity, trainingPasses,
phase="Training")
print "\nRunning test phase..."
experiment.runNetworkOnSequences(generatedSequences,
labeledSequences,
tmLearn=False,
upLearn=False,
verbosity=consoleVerbosity,
progressInterval=_SHOW_PROGRESS_INTERVAL)
print "\nPass\tBursting Columns Mean\tStdDev\tMax"
stats = experiment.getBurstingColumnsStats()
print "{0}\t{1}\t{2}\t{3}".format(0, stats[0], stats[1], stats[2])
if trainingPasses > 0 and stats[0] > 0:
print "***WARNING! MEAN BURSTING COLUMNS IN TEST PHASE IS GREATER THAN 0***"
print
print MonitorMixinBase.mmPrettyPrintMetrics(
experiment.tm.mmGetDefaultMetrics() + experiment.up.mmGetDefaultMetrics())
print
plotNetworkState(experiment, plotVerbosity, trainingPasses, phase="Testing")
elapsed = int(time.time() - start)
print "Total time: {0:2} seconds.".format(elapsed)
# Write Union SDR trace
metricName = "activeCells"
outputFileName = "unionSdrTrace_{0}learningPasses.csv".format(trainingPasses)
writeMetricTrace(experiment, metricName, outputDir, outputFileName)
if plotVerbosity >= 1:
raw_input("Press any key to exit...")
patternCardinality = params["patternCardinality"]
# Length of sequences shown to network
sequenceLength = params["sequenceLength"]
# Number of sequences used. Sequences may share common elements.
numberOfSequences = params["numberOfSequences"]
# Number of sequence passes for training the TM. Zero => no training.
trainingPasses = params["trainingPasses"]
# Generate a sequence list and an associated labeled list (both containing a
# set of sequences separated by None)
print "\nGenerating sequences..."
patternAlphabetSize = sequenceLength * numberOfSequences
patternMachine = PatternMachine(patternDimensionality, patternCardinality,
patternAlphabetSize)
sequenceMachine = SequenceMachine(patternMachine)
numbers = sequenceMachine.generateNumbers(numberOfSequences, sequenceLength)
generatedSequences = sequenceMachine.generateFromNumbers(numbers)
sequenceLabels = [
str(numbers[i + i * sequenceLength:i + (i + 1) * sequenceLength])
for i in xrange(numberOfSequences)
]
labeledSequences = []
for label in sequenceLabels:
for _ in xrange(sequenceLength):
labeledSequences.append(label)
labeledSequences.append(None)
def getPatternMachine(self):
return PatternMachine(self.n, self.w, num=300)
class ExtensiveColumnPoolerTest(unittest.TestCase):
"""
Algorithmic tests for the ColumnPooler region.
Each test actually tests multiple aspects of the algorithm. For more
atomic tests refer to column_pooler_unit_test.
The notation for objects is the following:
object{patternA, patternB, ...}
In these tests, the proximally-fed SDR's are simulated as unique (location,
feature) pairs regardless of actual locations and features, unless stated
otherwise.
"""
inputWidth = 2048 * 8
numInputActiveBits = int(0.02 * inputWidth)
outputWidth = 2048
numOutputActiveBits = 40
seed = 42
def testNewInputs(self):
"""
Checks that the behavior is correct when facing unseed inputs.
"""
self.init()
# feed the first input, a random SDR should be generated
initialPattern = self.generateObject(1)
self.learn(initialPattern, numRepetitions=1, newObject=True)
representation = self._getActiveRepresentation()
self.assertEqual(
len(representation),
self.numOutputActiveBits,
"The generated representation is incorrect"
)
# feed a new input for the same object, the previous SDR should persist
newPattern = self.generateObject(1)
self.learn(newPattern, numRepetitions=1, newObject=False)
newRepresentation = self._getActiveRepresentation()
self.assertNotEqual(initialPattern, newPattern)
self.assertEqual(
newRepresentation,
representation,
"The SDR did not persist when learning the same object"
)
# without sensory input, the SDR should persist as well
emptyPattern = [set()]
self.learn(emptyPattern, numRepetitions=1, newObject=False)
newRepresentation = self._getActiveRepresentation()
self.assertEqual(
newRepresentation,
representation,
"The SDR did not persist after an empty input."
)
def testLearnSinglePattern(self):
"""
A single pattern is learnt for a single object.
Objects: A{X, Y}
"""
self.init()
object = self.generateObject(1)
self.learn(object, numRepetitions=1, newObject=True)
# check that the active representation is sparse
representation = self._getActiveRepresentation()
self.assertEqual(
len(representation),
self.numOutputActiveBits,
"The generated representation is incorrect"
)
# check that the pattern was correctly learnt
self.infer(feedforwardPattern=object[0])
self.assertEqual(
self._getActiveRepresentation(),
representation,
"The pooled representation is not stable"
)
# present new pattern for same object
# it should be mapped to the same representation
newPattern = [self.generatePattern()]
self.learn(newPattern, numRepetitions=1, newObject=False)
# check that the active representation is sparse
newRepresentation = self._getActiveRepresentation()
self.assertEqual(
newRepresentation,
representation,
"The new pattern did not map to the same object representation"
)
# check that the pattern was correctly learnt and is stable
self.infer(feedforwardPattern=object[0])
self.assertEqual(
self._getActiveRepresentation(),
representation,
"The pooled representation is not stable"
)
def testLearnSingleObject(self):
"""
Many patterns are learnt for a single object.
Objects: A{P, Q, R, S, T}
"""
self.init()
object = self.generateObject(numPatterns=5)
self.learn(object, numRepetitions=1, randomOrder=True, newObject=True)
representation = self._getActiveRepresentation()
# check that all patterns map to the same object
for pattern in object:
self.infer(feedforwardPattern=pattern)
self.assertEqual(
self._getActiveRepresentation(),
representation,
"The pooled representation is not stable"
)
# if activity stops, check that the representation persists
self.infer(feedforwardPattern=set())
self.assertEqual(
self._getActiveRepresentation(),
representation,
"The pooled representation did not persist"
)
def testLearnTwoObjectNoCommonPattern(self):
"""
Same test as before, using two objects, without common pattern.
Objects: A{P, Q, R, S,T} B{V, W, X, Y, Z}
"""
self.init()
objectA = self.generateObject(numPatterns=5)
self.learn(objectA, numRepetitions=3, randomOrder=True, newObject=True)
representationA = self._getActiveRepresentation()
objectB = self.generateObject(numPatterns=5)
self.learn(objectB, numRepetitions=3, randomOrder=True, newObject=True)
representationB = self._getActiveRepresentation()
self.assertNotEqual(representationA, representationB)
# check that all patterns map to the same object
for pattern in objectA:
self.infer(feedforwardPattern=pattern)
self.assertEqual(
self._getActiveRepresentation(),
representationA,
"The pooled representation for the first object is not stable"
)
# check that all patterns map to the same object
for pattern in objectB:
self.infer(feedforwardPattern=pattern)
self.assertEqual(
self._getActiveRepresentation(),
representationB,
"The pooled representation for the second object is not stable"
)
# feed union of patterns in object A
pattern = objectA[0] | objectA[1]
self.infer(feedforwardPattern=pattern)
self.assertEqual(
self._getActiveRepresentation(),
representationA,
"The active representation is incorrect"
)
# feed unions of patterns in objects A and B
pattern = objectA[0] | objectB[0]
self.infer(feedforwardPattern=pattern)
self.assertEqual(
self._getActiveRepresentation(),
representationA | representationB,
"The active representation is incorrect"
)
def testLearnTwoObjectsOneCommonPattern(self):
"""
Same test as before, except the two objects share a pattern
Objects: A{P, Q, R, S,T} B{P, W, X, Y, Z}
"""
self.init()
objectA = self.generateObject(numPatterns=5)
self.learn(objectA, numRepetitions=3, randomOrder=True, newObject=True)
representationA = self._getActiveRepresentation()
objectB = self.generateObject(numPatterns=5)
objectB[0] = objectA[0]
self.learn(objectB, numRepetitions=3, randomOrder=True, newObject=True)
representationB = self._getActiveRepresentation()
self.assertNotEqual(representationA, representationB)
# very small overlap
self.assertLessEqual(len(representationA & representationB), 3)
# check that all patterns except the common one map to the same object
for pattern in objectA[1:]:
self.infer(feedforwardPattern=pattern)
self.assertEqual(
self._getActiveRepresentation(),
representationA,
"The pooled representation for the first object is not stable"
)
# check that all patterns except the common one map to the same object
for pattern in objectB[1:]:
self.infer(feedforwardPattern=pattern)
self.assertEqual(
self._getActiveRepresentation(),
representationB,
"The pooled representation for the second object is not stable"
)
# feed shared pattern
pattern = objectA[0]
self.infer(feedforwardPattern=pattern)
self.assertEqual(
self._getActiveRepresentation(),
representationA | representationB,
"The active representation is incorrect"
)
# feed union of patterns in object A
pattern = objectA[1] | objectA[2]
self.infer(feedforwardPattern=pattern)
self.assertEqual(
self._getActiveRepresentation(),
representationA,
"The active representation is incorrect"
)
# feed unions of patterns in objects A and B
pattern = objectA[1] | objectB[1]
self.infer(feedforwardPattern=pattern)
self.assertEqual(
self._getActiveRepresentation(),
representationA | representationB,
"The active representation is incorrect"
)
def testLearnThreeObjectsOneCommonPattern(self):
"""
Same test as before, with three objects
Objects: A{P, Q, R, S,T} B{P, W, X, Y, Z} C{W, H, I, K, L}
"""
self.init()
objectA = self.generateObject(numPatterns=5)
self.learn(objectA, numRepetitions=3, randomOrder=True, newObject=True)
representationA = self._getActiveRepresentation()
objectB = self.generateObject(numPatterns=5)
objectB[0] = objectA[0]
self.learn(objectB, numRepetitions=3, randomOrder=True, newObject=True)
representationB = self._getActiveRepresentation()
objectC = self.generateObject(numPatterns=5)
objectC[0] = objectB[1]
self.learn(objectC, numRepetitions=3, randomOrder=True, newObject=True)
representationC = self._getActiveRepresentation()
self.assertNotEquals(representationA, representationB, representationC)
# very small overlap
self.assertLessEqual(len(representationA & representationB), 3)
self.assertLessEqual(len(representationB & representationC), 3)
self.assertLessEqual(len(representationA & representationC), 3)
# check that all patterns except the common one map to the same object
for pattern in objectA[1:]:
self.infer(feedforwardPattern=pattern)
self.assertEqual(
self._getActiveRepresentation(),
representationA,
"The pooled representation for the first object is not stable"
)
# check that all patterns except the common one map to the same object
for pattern in objectB[2:]:
self.infer(feedforwardPattern=pattern)
self.assertEqual(
self._getActiveRepresentation(),
representationB,
"The pooled representation for the second object is not stable"
)
# check that all patterns except the common one map to the same object
for pattern in objectC[1:]:
self.infer(feedforwardPattern=pattern)
self.assertEqual(
self._getActiveRepresentation(),
representationC,
"The pooled representation for the third object is not stable"
)
# feed shared pattern between A and B
pattern = objectA[0]
self.infer(feedforwardPattern=pattern)
self.assertEqual(
self._getActiveRepresentation(),
representationA | representationB,
"The active representation is incorrect"
)
# feed shared pattern between B and C
pattern = objectB[1]
self.infer(feedforwardPattern=pattern)
self.assertEqual(
self._getActiveRepresentation(),
representationB | representationC,
"The active representation is incorrect"
)
# feed union of patterns in object A
pattern = objectA[1] | objectA[2]
self.infer(feedforwardPattern=pattern)
self.assertEqual(
self._getActiveRepresentation(),
representationA,
"The active representation is incorrect"
)
# feed unions of patterns to activate all objects
pattern = objectA[1] | objectB[1]
self.infer(feedforwardPattern=pattern)
self.assertEqual(
self._getActiveRepresentation(),
representationA | representationB | representationC,
"The active representation is incorrect"
)
def testLearnThreeObjectsOneCommonPatternSpatialNoise(self):
"""
Same test as before, with three objects
Objects: A{P, Q, R, S,T} B{P, W, X, Y, Z} C{W, H, I, K, L}
"""
self.init()
objectA = self.generateObject(numPatterns=5)
self.learn(objectA, numRepetitions=3, randomOrder=True, newObject=True)
representationA = self._getActiveRepresentation()
objectB = self.generateObject(numPatterns=5)
objectB[0] = objectA[0]
self.learn(objectB, numRepetitions=3, randomOrder=True, newObject=True)
representationB = self._getActiveRepresentation()
objectC = self.generateObject(numPatterns=5)
objectC[0] = objectB[1]
self.learn(objectC, numRepetitions=3, randomOrder=True, newObject=True)
representationC = self._getActiveRepresentation()
self.assertNotEquals(representationA, representationB, representationC)
# very small overlap
self.assertLessEqual(len(representationA & representationB), 3)
self.assertLessEqual(len(representationB & representationC), 3)
self.assertLessEqual(len(representationA & representationC), 3)
# check that all patterns except the common one map to the same object
for pattern in objectA[1:]:
noisyPattern = self.proximalPatternMachine.addNoise(pattern, 0.05)
self.infer(feedforwardPattern=noisyPattern)
self.assertEqual(
self._getActiveRepresentation(),
representationA,
"The pooled representation for the first object is not stable"
)
# check that all patterns except the common one map to the same object
for pattern in objectB[2:]:
noisyPattern = self.proximalPatternMachine.addNoise(pattern, 0.05)
self.infer(feedforwardPattern=noisyPattern)
self.assertEqual(
self._getActiveRepresentation(),
representationB,
"The pooled representation for the second object is not stable"
)
# check that all patterns except the common one map to the same object
for pattern in objectC[1:]:
noisyPattern = self.proximalPatternMachine.addNoise(pattern, 0.05)
self.infer(feedforwardPattern=noisyPattern)
self.assertEqual(
self._getActiveRepresentation(),
representationC,
"The pooled representation for the third object is not stable"
)
# feed shared pattern between A and B
pattern = objectA[0]
noisyPattern = self.proximalPatternMachine.addNoise(pattern, 0.05)
self.infer(feedforwardPattern=noisyPattern)
self.assertEqual(
self._getActiveRepresentation(),
representationA | representationB,
"The active representation is incorrect"
)
# feed shared pattern between B and C
pattern = objectB[1]
noisyPattern = self.proximalPatternMachine.addNoise(pattern, 0.05)
self.infer(feedforwardPattern=noisyPattern)
self.assertEqual(
self._getActiveRepresentation(),
representationB | representationC,
"The active representation is incorrect"
)
# feed union of patterns in object A
pattern = objectA[1] | objectA[2]
noisyPattern = self.proximalPatternMachine.addNoise(pattern, 0.05)
self.infer(feedforwardPattern=noisyPattern)
self.assertEqual(
self._getActiveRepresentation(),
representationA,
"The active representation is incorrect"
)
# feed unions of patterns to activate all objects
pattern = objectA[1] | objectB[1]
noisyPattern = self.proximalPatternMachine.addNoise(pattern, 0.05)
self.infer(feedforwardPattern=noisyPattern)
self.assertEqual(
self._getActiveRepresentation(),
representationA | representationB | representationC,
"The active representation is incorrect"
)
def testLearnOneObjectInTwoColumns(self):
"""Learns one object in two different columns."""
self.init(numCols=2)
neighborsIndices = [[1], [0]]
objectA = self.generateObject(numPatterns=5, numCols=2)
# learn object
self.learnMultipleColumns(
objectA,
numRepetitions=3,
neighborsIndices=neighborsIndices,
randomOrder=True,
newObject=True
)
objectARepresentations = self._getActiveRepresentations()
for pooler in self.poolers:
pooler.reset()
for patterns in objectA:
for i in xrange(3):
activeRepresentations = self._getActiveRepresentations()
self.inferMultipleColumns(
feedforwardPatterns=patterns,
activeRepresentations=activeRepresentations,
neighborsIndices=neighborsIndices,
)
if i > 0:
self.assertEqual(activeRepresentations,
self._getActiveRepresentations())
self.assertEqual(objectARepresentations,
self._getActiveRepresentations())
def testLearnTwoObjectsInTwoColumnsNoCommonPattern(self):
"""Learns two objects in two different columns."""
self.init(numCols=2)
neighborsIndices = [[1], [0]]
objectA = self.generateObject(numPatterns=5, numCols=2)
objectB = self.generateObject(numPatterns=5, numCols=2)
# learn object
self.learnMultipleColumns(
objectA,
numRepetitions=3,
neighborsIndices=neighborsIndices,
randomOrder=True,
newObject=True
)
activeRepresentationsA = self._getActiveRepresentations()
# learn object
self.learnMultipleColumns(
objectB,
numRepetitions=3,
neighborsIndices=neighborsIndices,
randomOrder=True,
newObject=True,
)
activeRepresentationsB = self._getActiveRepresentations()
for pooler in self.poolers:
pooler.reset()
# check inference for object A
# for the first pattern, the distal predictions won't be correct
firstPattern = True
for patternsA in objectA:
for i in xrange(3):
activeRepresentations = self._getActiveRepresentations()
self.inferMultipleColumns(
feedforwardPatterns=patternsA,
activeRepresentations=activeRepresentations,
neighborsIndices=neighborsIndices,
)
if firstPattern:
firstPattern = False
else:
self.assertEqual(
activeRepresentationsA,
self._getPredictedActiveCells()
)
self.assertEqual(
activeRepresentationsA,
self._getActiveRepresentations()
)
for pooler in self.poolers:
pooler.reset()
# check inference for object B
firstPattern = True
for patternsB in objectB:
for i in xrange(3):
activeRepresentations = self._getActiveRepresentations()
self.inferMultipleColumns(
feedforwardPatterns=patternsB,
activeRepresentations=activeRepresentations,
neighborsIndices=neighborsIndices
)
if firstPattern:
firstPattern = False
else:
self.assertEqual(
activeRepresentationsB,
self._getPredictedActiveCells()
)
self.assertEqual(
activeRepresentationsB,
self._getActiveRepresentations()
)
def testLearnTwoObjectsInTwoColumnsOneCommonPattern(self):
"""Learns two objects in two different columns, with a common pattern."""
self.init(numCols=2)
neighborsIndices = [[1], [0]]
objectA = self.generateObject(numPatterns=5, numCols=2)
objectB = self.generateObject(numPatterns=5, numCols=2)
# second pattern in column 0 is shared
objectB[1][0] = objectA[1][0]
# learn object
self.learnMultipleColumns(
objectA,
numRepetitions=3,
neighborsIndices=neighborsIndices,
randomOrder=True,
newObject=True
)
activeRepresentationsA = self._getActiveRepresentations()
# learn object
self.learnMultipleColumns(
objectB,
numRepetitions=3,
neighborsIndices=neighborsIndices,
randomOrder=True,
newObject=True
)
activeRepresentationsB = self._getActiveRepresentations()
# check inference for object A
# for the first pattern, the distal predictions won't be correct
# for the second one, the prediction will be unique thanks to the
# distal predictions from the other column which has no ambiguity
for pooler in self.poolers:
pooler.reset()
firstPattern = True
for patternsA in objectA:
for i in xrange(3):
activeRepresentations = self._getActiveRepresentations()
self.inferMultipleColumns(
feedforwardPatterns=patternsA,
activeRepresentations=activeRepresentations,
neighborsIndices=neighborsIndices,
)
if firstPattern:
firstPattern = False
else:
self.assertEqual(
activeRepresentationsA,
self._getPredictedActiveCells()
)
self.assertEqual(
activeRepresentationsA,
self._getActiveRepresentations()
)
for pooler in self.poolers:
pooler.reset()
# check inference for object B
firstPattern = True
for patternsB in objectB:
for i in xrange(3):
activeRepresentations = self._getActiveRepresentations()
self.inferMultipleColumns(
feedforwardPatterns=patternsB,
activeRepresentations=activeRepresentations,
neighborsIndices=neighborsIndices
)
if firstPattern:
firstPattern = False
else:
self.assertEqual(
activeRepresentationsB,
self._getPredictedActiveCells()
)
self.assertEqual(
activeRepresentationsB,
self._getActiveRepresentations()
)
def testLearnTwoObjectsInTwoColumnsOneCommonPatternEmptyFirstInput(self):
"""Learns two objects in two different columns, with a common pattern."""
self.init(numCols=2)
neighborsIndices = [[1], [0]]
objectA = self.generateObject(numPatterns=5, numCols=2)
objectB = self.generateObject(numPatterns=5, numCols=2)
# second pattern in column 0 is shared
objectB[1][0] = objectA[1][0]
# learn object
self.learnMultipleColumns(
objectA,
numRepetitions=3,
neighborsIndices=neighborsIndices,
randomOrder=True,
newObject=True
)
activeRepresentationsA = self._getActiveRepresentations()
# learn object
self.learnMultipleColumns(
objectB,
numRepetitions=3,
neighborsIndices=neighborsIndices,
randomOrder=True,
newObject=True
)
activeRepresentationsB = self._getActiveRepresentations()
# check inference for object A
for pooler in self.poolers:
pooler.reset()
firstPattern = True
for patternsA in objectA:
activeRepresentations = self._getActiveRepresentations()
if firstPattern:
self.inferMultipleColumns(
feedforwardPatterns=[set(), patternsA[1]],
activeRepresentations=activeRepresentations,
neighborsIndices=neighborsIndices,
)
desiredRepresentation = [set(), activeRepresentationsA[1]]
else:
self.inferMultipleColumns(
feedforwardPatterns=patternsA,
activeRepresentations=activeRepresentations,
neighborsIndices=neighborsIndices,
)
desiredRepresentation = activeRepresentationsA
self.assertEqual(
desiredRepresentation,
self._getActiveRepresentations()
)
def testPersistence(self):
"""After learning, representation should persist in L2 without input."""
self.init(numCols=2)
neighborsIndices = [[1], [0]]
objectA = self.generateObject(numPatterns=5, numCols=2)
# learn object
self.learnMultipleColumns(
objectA,
numRepetitions=3,
neighborsIndices=neighborsIndices,
randomOrder=True,
newObject=True
)
objectARepresentations = self._getActiveRepresentations()
for pooler in self.poolers:
pooler.reset()
for patterns in objectA:
for i in xrange(3):
# replace third pattern for column 2 by empty pattern
if i == 2:
patterns[1] = set()
activeRepresentations = self._getActiveRepresentations()
self.inferMultipleColumns(
feedforwardPatterns=patterns,
activeRepresentations=activeRepresentations,
neighborsIndices=neighborsIndices,
)
if i > 0:
self.assertEqual(activeRepresentations,
self._getActiveRepresentations())
self.assertEqual(objectARepresentations,
self._getActiveRepresentations())
def testLateralDisambiguation(self):
"""Lateral disambiguation using a constant simulated distal input."""
self.init()
objectA = self.generateObject(numPatterns=5)
lateralInputA = [None] + [self.generatePattern() for _ in xrange(4)]
self.learn(objectA,
lateralPatterns=lateralInputA,
numRepetitions=3,
randomOrder=True,
newObject=True)
representationA = self._getActiveRepresentation()
objectB = self.generateObject(numPatterns=5)
objectB[3] = objectA[3]
lateralInputB = [None] + [self.generatePattern() for _ in xrange(4)]
self.learn(objectB,
lateralPatterns=lateralInputB,
numRepetitions=3,
randomOrder=True,
newObject=True)
representationB = self._getActiveRepresentation()
self.assertNotEqual(representationA, representationB)
# very small overlap
self.assertLessEqual(len(representationA & representationB), 3)
# no ambiguity with lateral input
for pattern in objectA:
self.infer(feedforwardPattern=pattern, lateralInput=lateralInputA[-1])
self.assertEqual(
self._getActiveRepresentation(),
representationA,
"The pooled representation for the first object is not stable"
)
# no ambiguity with lateral input
for pattern in objectB:
self.infer(feedforwardPattern=pattern, lateralInput=lateralInputB[-1])
self.assertEqual(
self._getActiveRepresentation(),
representationB,
"The pooled representation for the second object is not stable"
)
@unittest.skip("Fails, need to discuss")
def testMultiColumnCompetition(self):
"""Competition between multiple conflicting lateral inputs."""
self.init(numCols=4)
neighborsIndices = [[1, 2, 3], [0, 2, 3], [0, 1, 3], [0, 1, 2]]
objectA = self.generateObject(numPatterns=5, numCols=4)
objectB = self.generateObject(numPatterns=5, numCols=4)
# second pattern in column 0 is shared
objectB[1][0] = objectA[1][0]
# learn object
self.learnMultipleColumns(
objectA,
numRepetitions=3,
neighborsIndices=neighborsIndices,
randomOrder=True,
newObject=True
)
activeRepresentationsA = self._getActiveRepresentations()
# learn object
self.learnMultipleColumns(
objectB,
numRepetitions=3,
neighborsIndices=neighborsIndices,
randomOrder=True,
newObject=True
)
activeRepresentationsB = self._getActiveRepresentations()
# check inference for object A
# for the first pattern, the distal predictions won't be correct
# for the second one, the prediction will be unique thanks to the
# distal predictions from the other column which has no ambiguity
for pooler in self.poolers:
pooler.reset()
# sensed patterns will be mixed
sensedPatterns = objectA[1][:-1] + [objectA[1][-1] | objectB[1][-1]]
# feed sensed patterns first time
# every one feels the correct object, except first column which feels
# the union (reminder: lateral input are delayed)
activeRepresentations = self._getActiveRepresentations()
self.inferMultipleColumns(
feedforwardPatterns=sensedPatterns,
activeRepresentations=activeRepresentations,
neighborsIndices=neighborsIndices,
)
firstSensedRepresentations = [
activeRepresentationsA[0] | activeRepresentationsB[0],
activeRepresentationsA[1],
activeRepresentationsA[2],
activeRepresentationsA[3] | activeRepresentationsB[3]
]
self.assertEqual(
firstSensedRepresentations,
self._getActiveRepresentations()
)
# feed sensed patterns second time
# the distal predictions are still ambiguous in C1, but disambiguated
# in C4
activeRepresentations = self._getActiveRepresentations()
self.inferMultipleColumns(
feedforwardPatterns=sensedPatterns,
activeRepresentations=activeRepresentations,
neighborsIndices=neighborsIndices,
)
secondSensedRepresentations = [
activeRepresentationsA[0] | activeRepresentationsB[0],
activeRepresentationsA[1],
activeRepresentationsA[2],
activeRepresentationsA[3]
]
self.assertEqual(
secondSensedRepresentations,
self._getActiveRepresentations()
)
# feed sensed patterns third time
# this time, it is all disambiguated
activeRepresentations = self._getActiveRepresentations()
self.inferMultipleColumns(
feedforwardPatterns=sensedPatterns,
activeRepresentations=activeRepresentations,
neighborsIndices=neighborsIndices,
)
self.assertEqual(
activeRepresentationsA,
self._getActiveRepresentations()
)
def testMutualDisambiguationThroughUnions(self):
"""
Learns three object in two different columns.
Feed ambiguous sensations, A u B and B u C. The system should narrow down
to B.
"""
self.init(numCols=2)
neighborsIndices = [[1], [0]]
objectA = self.generateObject(numPatterns=5, numCols=2)
objectB = self.generateObject(numPatterns=5, numCols=2)
objectC = self.generateObject(numPatterns=5, numCols=2)
# learn object
self.learnMultipleColumns(
objectA,
numRepetitions=3,
neighborsIndices=neighborsIndices,
randomOrder=True,
newObject=True
)
activeRepresentationsA = self._getActiveRepresentations()
# learn object
self.learnMultipleColumns(
objectB,
numRepetitions=3,
neighborsIndices=neighborsIndices,
randomOrder=True,
newObject=True
)
activeRepresentationsB = self._getActiveRepresentations()
# learn object
self.learnMultipleColumns(
objectC,
numRepetitions=3,
neighborsIndices=neighborsIndices,
randomOrder=True,
newObject=True
)
activeRepresentationsC = self._getActiveRepresentations()
# create sensed patterns (ambiguous)
sensedPatterns = [objectA[1][0] | objectB[1][0],
objectB[2][1] | objectC[2][1]]
for pooler in self.poolers:
pooler.reset()
# feed sensed patterns first time
# the L2 representations should be ambiguous
activeRepresentations = self._getActiveRepresentations()
self.inferMultipleColumns(
feedforwardPatterns=sensedPatterns,
activeRepresentations=activeRepresentations,
neighborsIndices=neighborsIndices,
)
firstRepresentations = [
activeRepresentationsA[0] | activeRepresentationsB[0],
activeRepresentationsB[1] | activeRepresentationsC[1]
]
self.assertEqual(
firstRepresentations,
self._getActiveRepresentations()
)
# feed a second time
# the L2 representations should be ambiguous
activeRepresentations = self._getActiveRepresentations()
self.inferMultipleColumns(
feedforwardPatterns=sensedPatterns,
activeRepresentations=activeRepresentations,
neighborsIndices=neighborsIndices,
)
self.assertEqual(
firstRepresentations,
self._getActiveRepresentations()
)
# feed a third time, distal predictions should disambiguate
# we are using the third time because there is an off-by-one in pooler
activeRepresentations = self._getActiveRepresentations()
self.inferMultipleColumns(
feedforwardPatterns=sensedPatterns,
activeRepresentations=activeRepresentations,
neighborsIndices=neighborsIndices,
)
# check that representations are unique, being slightly tolerant
self.assertLessEqual(
len(self._getActiveRepresentations()[0] - activeRepresentationsB[0]),
5,
)
self.assertLessEqual(
len(self._getActiveRepresentations()[1] - activeRepresentationsB[1]),
5,
)
self.assertGreaterEqual(
len(self._getActiveRepresentations()[0] & activeRepresentationsB[0]),
35,
)
self.assertGreaterEqual(
len(self._getActiveRepresentations()[1] & activeRepresentationsB[1]),
35,
)
self.assertEqual(
self._getActiveRepresentations(),
self._getPredictedActiveCells(),
)
def setUp(self):
"""
Sets up the test.
"""
# single column case
self.pooler = None
# multi column case
self.poolers = []
# create pattern machine
self.proximalPatternMachine = PatternMachine(
n=self.inputWidth,
w=self.numOutputActiveBits,
num=200,
seed=self.seed
)
self.patternId = 0
np.random.seed(self.seed)
# Wrappers around ColumnPooler API
def learn(self,
feedforwardPatterns,
lateralPatterns=None,
numRepetitions=1,
randomOrder=True,
newObject=True):
"""
Parameters:
----------------------------
Learns a single object, with the provided patterns.
@param feedforwardPatterns (list(set))
List of proximal input patterns
@param lateralPatterns (list(list(set)))
List of distal input patterns, or None. If no lateral input is
used. The outer list is expected to have the same length as
feedforwardPatterns, whereas each inner list's length is the
number of cortical columns which are distally connected to the
pooler.
@param numRepetitions (int)
Number of times the patterns will be fed
@param randomOrder (bool)
If true, the order of patterns will be shuffled at each
repetition
"""
if newObject:
self.pooler.mmClearHistory()
self.pooler.reset()
# set-up
indices = range(len(feedforwardPatterns))
if lateralPatterns is None:
lateralPatterns = [None] * len(feedforwardPatterns)
for _ in xrange(numRepetitions):
if randomOrder:
np.random.shuffle(indices)
for idx in indices:
self.pooler.compute(feedforwardPatterns[idx],
activeExternalCells=lateralPatterns[idx],
learn=True)
def infer(self,
feedforwardPattern,
lateralInput=None,
printMetrics=False):
"""
Feeds a single pattern to the column pooler (as well as an eventual lateral
pattern).
Parameters:
----------------------------
@param feedforwardPattern (set)
Input proximal pattern to the pooler
@param lateralPatterns (list(set))
Input dislal patterns to the pooler (one for each neighboring CC's)
@param printMetrics (bool)
If true, will print cell metrics
"""
self.pooler.compute(feedforwardPattern,
activeExternalCells=lateralInput,
learn=False)
if printMetrics:
print self.pooler.mmPrettyPrintMetrics(
self.pooler.mmGetDefaultMetrics()
)
# Helper functions
def generatePattern(self):
"""
Returns a random proximal input pattern.
"""
pattern = self.proximalPatternMachine.get(self.patternId)
self.patternId += 1
return pattern
def generateObject(self, numPatterns, numCols=1):
"""
Creates a list of patterns, for a given object.
If numCols > 1 is given, a list of list of patterns will be returned.
"""
if numCols == 1:
return [self.generatePattern() for _ in xrange(numPatterns)]
else:
patterns = []
for i in xrange(numPatterns):
patterns.append([self.generatePattern() for _ in xrange(numCols)])
return patterns
def init(self, overrides=None, numCols=1):
"""
Creates the column pooler with specified parameter overrides.
Except for the specified overrides and problem-specific parameters, used
parameters are implementation defaults.
"""
params = {
"inputWidth": self.inputWidth,
"numActiveColumnsPerInhArea": self.numOutputActiveBits,
"columnDimensions": (self.outputWidth,),
"seed": self.seed,
"initialPermanence": 0.51,
"connectedPermanence": 0.6,
"permanenceIncrement": 0.1,
"permanenceDecrement": 0.02,
"minThreshold": 10,
"predictedSegmentDecrement": 0.004,
"activationThreshold": 10,
"maxNewSynapseCount": 20,
"maxSegmentsPerCell": 255,
"maxSynapsesPerSegment": 255,
}
if overrides is None:
overrides = {}
params.update(overrides)
if numCols == 1:
self.pooler = MonitoredColumnPooler(**params)
else:
# TODO: We need a different seed for each pooler otherwise each one
# outputs an identical representation. Use random seed for now but ideally
# we would set different specific seeds for each pooler
params['seed']=0
self.poolers = [MonitoredColumnPooler(**params) for _ in xrange(numCols)]
def _getActiveRepresentation(self):
"""
Retrieves the current active representation in the pooler.
"""
if self.pooler is None:
raise ValueError("No pooler has been instantiated")
return set(self.pooler.getActiveCells())
# Multi-column testing
def learnMultipleColumns(self,
feedforwardPatterns,
numRepetitions=1,
neighborsIndices=None,
randomOrder=True,
newObject=True):
"""
Learns a single object, feeding it through the multiple columns.
Parameters:
----------------------------
Learns a single object, with the provided patterns.
@param feedforwardPatterns (list(list(set)))
List of proximal input patterns (one for each pooler).
@param neighborsIndices (list(list))
List of column indices each column received input from.
@param numRepetitions (int)
Number of times the patterns will be fed
@param randomOrder (bool)
If true, the order of patterns will be shuffled at each
repetition
"""
if newObject:
for pooler in self.poolers:
pooler.mmClearHistory()
pooler.reset()
# use different set of pattern indices to allow random orders
indices = [range(len(feedforwardPatterns))] * len(self.poolers)
representations = [set()] * len(self.poolers)
# by default, all columns are neighbors
if neighborsIndices is None:
neighborsIndices = [
range(i) + range(i+1, len(self.poolers))
for i in xrange(len(self.poolers))
]
for _ in xrange(numRepetitions):
# independently shuffle pattern orders if necessary
if randomOrder:
for idx in indices:
np.random.shuffle(idx)
for i in xrange(len(indices[0])):
# get union of relevant lateral representations
lateralInputs = []
for col in xrange(len(self.poolers)):
lateralInputsCol = set()
for idx in neighborsIndices[col]:
lateralInputsCol = lateralInputsCol.union(representations[idx])
lateralInputs.append(lateralInputsCol)
# Train each column
for col in xrange(len(self.poolers)):
self.poolers[col].compute(
feedforwardInput=feedforwardPatterns[indices[col][i]][col],
activeExternalCells=lateralInputs[col],
learn=True
)
# update active representations
representations = self._getActiveRepresentations()
for i in xrange(len(representations)):
representations[i] = set([i * self.outputWidth + k \
for k in representations[i]])
def inferMultipleColumns(self,
feedforwardPatterns,
activeRepresentations=None,
neighborsIndices=None,
printMetrics=False,
reset=False):
"""
Feeds a single pattern to the column pooler (as well as an eventual lateral
pattern).
Parameters:
----------------------------
@param feedforwardPattern (list(set))
Input proximal patterns to the pooler (one for each column)
@param activeRepresentations (list(set))
Active representations in the columns at the previous step.
@param neighborsIndices (list(list))
List of column indices each column received input from.
@param printMetrics (bool)
If true, will print cell metrics
"""
if reset:
for pooler in self.poolers:
pooler.reset()
# create copy of activeRepresentations to not mutate it
representations = [None] * len(self.poolers)
# by default, all columns are neighbors
if neighborsIndices is None:
neighborsIndices = [
range(i) + range(i+1, len(self.poolers))
for i in xrange(len(self.poolers))
]
for i in xrange(len(self.poolers)):
if activeRepresentations[i] is not None:
representations[i] = set(i * self.outputWidth + k \
for k in activeRepresentations[i])
for col in range(len(self.poolers)):
lateralInputs = [representations[idx] for idx in neighborsIndices[col]]
if len(lateralInputs) > 0:
lateralInputs = set.union(*lateralInputs)
else:
lateralInputs = set()
self.poolers[col].compute(
feedforwardPatterns[col],
activeExternalCells=lateralInputs,
learn=False
)
if printMetrics:
for pooler in self.poolers:
print pooler.mmPrettyPrintMetrics(
pooler.mmGetDefaultMetrics()
)
def _getActiveRepresentations(self):
"""
Retrieves the current active representations in the poolers.
"""
if len(self.poolers) == 0:
raise ValueError("No pooler has been instantiated")
return [set(pooler.getActiveCells()) for pooler in self.poolers]
def _getPredictedActiveCells(self):
"""
Retrieves the current active representations in the poolers.
"""
if len(self.poolers) == 0:
raise ValueError("No pooler has been instantiated")
return [set(pooler.getActiveCells()) & set(pooler.tm.getPredictiveCells())\
for pooler in self.poolers]
def experiment1():
paramDir = 'params/1024_baseline/5_trainingPasses.yaml'
outputDir = 'results/'
params = yaml.safe_load(open(paramDir, 'r'))
options = {'plotVerbosity': 2, 'consoleVerbosity': 2}
plotVerbosity = 2
consoleVerbosity = 1
print "Running SDR overlap experiment...\n"
print "Params dir: {0}".format(paramDir)
print "Output dir: {0}\n".format(outputDir)
# Dimensionality of sequence patterns
patternDimensionality = params["patternDimensionality"]
# Cardinality (ON / true bits) of sequence patterns
patternCardinality = params["patternCardinality"]
# TODO If this parameter is to be supported, the sequence generation code
# below must change
# Number of unique patterns from which sequences are built
# patternAlphabetSize = params["patternAlphabetSize"]
# Length of sequences shown to network
sequenceLength = params["sequenceLength"]
# Number of sequences used. Sequences may share common elements.
numberOfSequences = params["numberOfSequences"]
# Number of sequence passes for training the TM. Zero => no training.
trainingPasses = params["trainingPasses"]
tmParamOverrides = params["temporalMemoryParams"]
upParamOverrides = params["unionPoolerParams"]
# Generate a sequence list and an associated labeled list (both containing a
# set of sequences separated by None)
start = time.time()
print "\nGenerating sequences..."
patternAlphabetSize = sequenceLength * numberOfSequences
patternMachine = PatternMachine(patternDimensionality, patternCardinality,
patternAlphabetSize)
sequenceMachine = SequenceMachine(patternMachine)
numbers = sequenceMachine.generateNumbers(numberOfSequences,
sequenceLength)
generatedSequences = sequenceMachine.generateFromNumbers(numbers)
sequenceLabels = [
str(numbers[i + i * sequenceLength:i + (i + 1) * sequenceLength])
for i in xrange(numberOfSequences)
]
labeledSequences = []
for label in sequenceLabels:
for _ in xrange(sequenceLength):
labeledSequences.append(label)
labeledSequences.append(None)
# Set up the Temporal Memory and Union Pooler network
print "\nCreating network..."
experiment = UnionTemporalPoolerExperiment(tmParamOverrides,
upParamOverrides)
# Train only the Temporal Memory on the generated sequences
if trainingPasses > 0:
print "\nTraining Temporal Memory..."
if consoleVerbosity > 0:
print "\nPass\tBursting Columns Mean\tStdDev\tMax"
for i in xrange(trainingPasses):
experiment.runNetworkOnSequences(
generatedSequences,
labeledSequences,
tmLearn=True,
upLearn=None,
verbosity=consoleVerbosity,
progressInterval=_SHOW_PROGRESS_INTERVAL)
if consoleVerbosity > 0:
stats = experiment.getBurstingColumnsStats()
print "{0}\t{1}\t{2}\t{3}".format(i, stats[0], stats[1],
stats[2])
# Reset the TM monitor mixin's records accrued during this training pass
# experiment.tm.mmClearHistory()
print
print MonitorMixinBase.mmPrettyPrintMetrics(
experiment.tm.mmGetDefaultMetrics())
print
experiment.tm.mmClearHistory()
experiment.up.mmClearHistory()
print "\nRunning test phase..."
inputSequences = generatedSequences
inputCategories = labeledSequences
tmLearn = True
upLearn = False
classifierLearn = False
currentTime = time.time()
experiment.tm.reset()
experiment.up.reset()
poolingActivationTrace = numpy.zeros((experiment.up._numColumns, 1))
activeCellsTrace = numpy.zeros((experiment.up._numColumns, 1))
activeSPTrace = numpy.zeros((experiment.up._numColumns, 1))
for _ in xrange(trainingPasses):
experiment.tm.reset()
for i in xrange(len(inputSequences)):
sensorPattern = inputSequences[i]
inputCategory = inputCategories[i]
if sensorPattern is None:
pass
else:
experiment.tm.compute(sensorPattern,
learn=tmLearn,
sequenceLabel=inputCategory)
if upLearn is not None:
activeCells, predActiveCells, burstingCols, = experiment.getUnionTemporalPoolerInput(
)
experiment.up.compute(activeCells,
predActiveCells,
learn=upLearn,
sequenceLabel=inputCategory)
currentPoolingActivation = experiment.up._poolingActivation
currentPoolingActivation = experiment.up._poolingActivation.reshape(
(experiment.up._numColumns, 1))
poolingActivationTrace = numpy.concatenate(
(poolingActivationTrace, currentPoolingActivation), 1)
currentUnionSDR = numpy.zeros(
(experiment.up._numColumns, 1))
currentUnionSDR[experiment.up._unionSDR] = 1
activeCellsTrace = numpy.concatenate(
(activeCellsTrace, currentUnionSDR), 1)
currentSPSDR = numpy.zeros((experiment.up._numColumns, 1))
currentSPSDR[experiment.up._activeCells] = 1
activeSPTrace = numpy.concatenate(
(activeSPTrace, currentSPSDR), 1)
print "\nPass\tBursting Columns Mean\tStdDev\tMax"
stats = experiment.getBurstingColumnsStats()
print "{0}\t{1}\t{2}\t{3}".format(0, stats[0], stats[1], stats[2])
print
print MonitorMixinBase.mmPrettyPrintMetrics(\
experiment.tm.mmGetDefaultMetrics() + experiment.up.mmGetDefaultMetrics())
print
experiment.tm.mmClearHistory()
# estimate fraction of shared bits across adjacent time point
unionSDRshared = experiment.up._mmComputeUnionSDRdiff()
bitLifeList = experiment.up._mmComputeBitLifeStats()
bitLife = numpy.array(bitLifeList)
# Plot SP outputs, UP persistence and UP outputs in testing phase
def showSequenceStartLine(ax, trainingPasses, sequenceLength):
for i in xrange(trainingPasses):
ax.vlines(i * sequenceLength, 0, 100, linestyles='--')
plt.figure()
ncolShow = 100
f, (ax1, ax2, ax3) = plt.subplots(nrows=1, ncols=3)
ax1.imshow(activeSPTrace[1:ncolShow, :],
cmap=cm.Greys,
interpolation="nearest",
aspect='auto')
showSequenceStartLine(ax1, trainingPasses, sequenceLength)
ax1.set_title('SP SDR')
ax1.set_ylabel('Columns')
ax2.imshow(poolingActivationTrace[1:100, :],
cmap=cm.Greys,
interpolation="nearest",
aspect='auto')
showSequenceStartLine(ax2, trainingPasses, sequenceLength)
ax2.set_title('Persistence')
ax3.imshow(activeCellsTrace[1:ncolShow, :],
cmap=cm.Greys,
interpolation="nearest",
aspect='auto')
showSequenceStartLine(ax3, trainingPasses, sequenceLength)
plt.title('Union SDR')
ax2.set_xlabel('Time (steps)')
pp = PdfPages('results/UnionPoolingOnLearnedTM_Experiment1.pdf')
pp.savefig()
pp.close()
f, (ax1, ax2, ax3) = plt.subplots(nrows=3, ncols=1)
ax1.plot((sum(activeCellsTrace)) / experiment.up._numColumns * 100)
ax1.set_ylabel('Union SDR size (%)')
ax1.set_xlabel('Time (steps)')
ax1.set_ylim(0, 25)
ax2.plot(unionSDRshared)
ax2.set_ylabel('Shared Bits')
ax2.set_xlabel('Time (steps)')
ax3.hist(bitLife)
ax3.set_xlabel('Life duration for each bit')
pp = PdfPages('results/UnionSDRproperty_Experiment1.pdf')
pp.savefig()
pp.close()
class ExtensiveTemporalMemoryTest(AbstractTemporalMemoryTest):
"""
==============================================================================
Basic First Order Sequences
==============================================================================
These tests ensure the most basic (first order) sequence learning mechanism is
working.
Parameters: Use a "fast learning mode": initPerm should be greater than
connectedPerm and permanenceDec should be zero. With these settings sequences
should be learned in one pass:
minThreshold = newSynapseCount
initialPermanence = 0.8
connectedPermanence = 0.7
permanenceDecrement = 0
permanenceIncrement = 0.4
Other Parameters:
columnDimensions = [100]
cellsPerColumn = 1
newSynapseCount = 11
activationThreshold = 11
Note: this is not a high order sequence, so one cell per column is fine.
Input Sequence: We train with M input sequences, each consisting of N random
patterns. Each pattern consists of a random number of bits on. The number of
1's in each pattern should be between 21 and 25 columns.
Each input pattern can optionally have an amount of spatial noise represented
by X, where X is the probability of switching an on bit with a random bit.
Training: The TP is trained with P passes of the M sequences. There
should be a reset between sequences. The total number of iterations during
training is P*N*M.
Testing: Run inference through the same set of sequences, with a reset before
each sequence. For each sequence the system should accurately predict the
pattern at the next time step up to and including the N-1'st pattern. The number
of predicted inactive cells at each time step should be reasonably low.
We can also calculate the number of synapses that should be
learned. We raise an error if too many or too few were learned.
B1) Basic sequence learner. M=1, N=100, P=1.
B2) Same as above, except P=2. Test that permanences go up and that no
additional synapses are learned. [TODO]
B3) N=300, M=1, P=1. (See how high we can go with N)
B4) N=100, M=3, P=1. (See how high we can go with N*M)
B5) Like B1 but with cellsPerColumn = 4. First order sequences should still
work just fine.
B6) Like B4 but with cellsPerColumn = 4. First order sequences should still
work just fine.
B7) Like B1 but with slower learning. Set the following parameters differently:
initialPermanence = 0.2
connectedPermanence = 0.7
permanenceIncrement = 0.2
Now we train the TP with the B1 sequence 4 times (P=4). This will increment
the permanences to be above 0.8 and at that point the inference will be correct.
This test will ensure the basic match function and segment activation rules are
working correctly.
B8) Like B7 but with 4 cells per column. Should still work.
B9) Like B7 but present the sequence less than 4 times: the inference should be
incorrect.
B10) Like B2, except that cells per column = 4. Should still add zero additional
synapses. [TODO]
B11) Like B5, but with activationThreshold = 8 and with each pattern
corrupted by a small amount of spatial noise (X = 0.05).
===============================================================================
High Order Sequences
===============================================================================
These tests ensure that high order sequences can be learned in a multiple cells
per column instantiation.
Parameters: Same as Basic First Order Tests above, but with varying cells per
column.
Input Sequence: We train with M input sequences, each consisting of N random
patterns. Each pattern consists of a random number of bits on. The number of
1's in each pattern should be between 21 and 25 columns. The sequences are
constructed to contain shared subsequences, such as:
A B C D E F G H I J
K L M D E F N O P Q
The position and length of shared subsequences are parameters in the tests.
Each input pattern can optionally have an amount of spatial noise represented
by X, where X is the probability of switching an on bit with a random bit.
Training: Identical to basic first order tests above.
Testing: Identical to basic first order tests above unless noted.
We can also calculate the number of segments and synapses that should be
learned. We raise an error if too many or too few were learned.
H1) Learn two sequences with a shared subsequence in the middle. Parameters
should be the same as B1. Since cellsPerColumn == 1, it should make more
predictions than necessary.
H2) Same as H1, but with cellsPerColumn == 4, and train multiple times.
It should make just the right number of predictions.
H3) Like H2, except the shared subsequence is in the beginning (e.g.
"ABCDEF" and "ABCGHIJ"). At the point where the shared subsequence ends, all
possible next patterns should be predicted. As soon as you see the first unique
pattern, the predictions should collapse to be a perfect prediction.
H4) Shared patterns. Similar to H2 except that patterns are shared between
sequences. All sequences are different shufflings of the same set of N
patterns (there is no shared subsequence).
H5) Combination of H4) and H2). Shared patterns in different sequences, with a
shared subsequence.
H6) Stress test: every other pattern is shared. [TODO]
H7) Start predicting in the middle of a sequence. [TODO]
H8) Hub capacity. How many patterns can use that hub? [TODO]
H9) Sensitivity to small amounts of spatial noise during inference (X = 0.05).
Parameters the same as B11, and sequences like H2.
H10) Higher order patterns with alternating elements.
Create the following 4 sequences:
A B A B A C
A B A B D E
A B F G H I
A J K L M N
After training we should verify that the expected transitions are in the
model. Prediction accuracy should be perfect. In addition, during inference,
after the first element is presented, the columns should not burst any more.
Need to verify, for the first sequence, that the high order representation
when presented with the second A and B is different from the representation
in the first presentation. [TODO]
"""
VERBOSITY = 1
DEFAULT_TM_PARAMS = {
"columnDimensions": [100],
"cellsPerColumn": 1,
"initialPermanence": 0.8,
"connectedPermanence": 0.7,
"minThreshold": 11,
"maxNewSynapseCount": 11,
"permanenceIncrement": 0.4,
"permanenceDecrement": 0,
"activationThreshold": 11
}
PATTERN_MACHINE = PatternMachine(100, range(21, 26), num=300)
def testB1(self):
"""Basic sequence learner. M=1, N=100, P=1."""
self.init()
numbers = self.sequenceMachine.generateNumbers(1, 100)
sequence = self.sequenceMachine.generateFromNumbers(numbers)
self.feedTM(sequence)
self._testTM(sequence)
self.assertAllActiveWerePredicted()
self.assertAllInactiveWereUnpredicted()
def testB3(self):
"""N=300, M=1, P=1. (See how high we can go with N)"""
self.init()
numbers = self.sequenceMachine.generateNumbers(1, 300)
sequence = self.sequenceMachine.generateFromNumbers(numbers)
self.feedTM(sequence)
self._testTM(sequence)
self.assertAllActiveWerePredicted()
self.assertAllInactiveWereUnpredicted()
def testB4(self):
"""N=100, M=3, P=1. (See how high we can go with N*M)"""
self.init()
numbers = self.sequenceMachine.generateNumbers(3, 100)
sequence = self.sequenceMachine.generateFromNumbers(numbers)
self.feedTM(sequence)
self._testTM(sequence)
self.assertAllActiveWerePredicted()
def testB5(self):
"""Like B1 but with cellsPerColumn = 4.
First order sequences should still work just fine."""
self.init({"cellsPerColumn": 4})
numbers = self.sequenceMachine.generateNumbers(1, 100)
sequence = self.sequenceMachine.generateFromNumbers(numbers)
self.feedTM(sequence)
self._testTM(sequence)
self.assertAllActiveWerePredicted()
self.assertAllInactiveWereUnpredicted()
def testB6(self):
"""Like B4 but with cellsPerColumn = 4.
First order sequences should still work just fine."""
self.init({"cellsPerColumn": 4})
numbers = self.sequenceMachine.generateNumbers(3, 100)
sequence = self.sequenceMachine.generateFromNumbers(numbers)
self.feedTM(sequence)
self._testTM(sequence)
self.assertAllActiveWerePredicted()
self.assertAllInactiveWereUnpredicted()
def testB7(self):
"""Like B1 but with slower learning.
Set the following parameters differently:
initialPermanence = 0.2
connectedPermanence = 0.7
permanenceIncrement = 0.2
Now we train the TP with the B1 sequence 4 times (P=4). This will increment
the permanences to be above 0.8 and at that point the inference will be correct.
This test will ensure the basic match function and segment activation rules are
working correctly.
"""
self.init({"initialPermanence": 0.2,
"connectedPermanence": 0.7,
"permanenceIncrement": 0.2})
numbers = self.sequenceMachine.generateNumbers(1, 100)
sequence = self.sequenceMachine.generateFromNumbers(numbers)
for _ in xrange(4):
self.feedTM(sequence)
self._testTM(sequence)
self.assertAllActiveWerePredicted()
self.assertAllInactiveWereUnpredicted()
def testB8(self):
"""Like B7 but with 4 cells per column.
Should still work."""
self.init({"initialPermanence": 0.2,
"connectedPermanence": 0.7,
"permanenceIncrement": 0.2,
"cellsPerColumn": 4})
numbers = self.sequenceMachine.generateNumbers(1, 100)
sequence = self.sequenceMachine.generateFromNumbers(numbers)
for _ in xrange(4):
self.feedTM(sequence)
self._testTM(sequence)
self.assertAllActiveWerePredicted()
self.assertAllInactiveWereUnpredicted()
def testB9(self):
"""Like B7 but present the sequence less than 4 times.
The inference should be incorrect."""
self.init({"initialPermanence": 0.2,
"connectedPermanence": 0.7,
"permanenceIncrement": 0.2})
numbers = self.sequenceMachine.generateNumbers(1, 100)
sequence = self.sequenceMachine.generateFromNumbers(numbers)
for _ in xrange(3):
self.feedTM(sequence)
self._testTM(sequence)
self.assertAllActiveWereUnpredicted()
def testB11(self):
"""Like B5, but with activationThreshold = 8 and with each pattern
corrupted by a small amount of spatial noise (X = 0.05)."""
self.init({"cellsPerColumn": 4,
"activationThreshold": 8})
numbers = self.sequenceMachine.generateNumbers(1, 100)
sequence = self.sequenceMachine.generateFromNumbers(numbers)
self.feedTM(sequence)
sequence = self.sequenceMachine.addSpatialNoise(sequence, 0.05)
self._testTM(sequence)
unpredictedActiveColumnsMetric = self.tm.mmGetMetricFromTrace(
self.tm.mmGetTraceUnpredictedActiveColumns())
self.assertTrue(unpredictedActiveColumnsMetric.mean < 1)
def testH1(self):
"""Learn two sequences with a short shared pattern.
Parameters should be the same as B1.
Since cellsPerColumn == 1, it should make more predictions than necessary.
"""
self.init()
numbers = self.sequenceMachine.generateNumbers(2, 20, (10, 15))
sequence = self.sequenceMachine.generateFromNumbers(numbers)
self.feedTM(sequence)
self._testTM(sequence)
self.assertAllActiveWerePredicted()
predictedInactiveColumnsMetric = self.tm.mmGetMetricFromTrace(
self.tm.mmGetTracePredictedInactiveColumns())
self.assertTrue(predictedInactiveColumnsMetric.mean > 0)
# At the end of both shared sequences, there should be
# predicted but inactive columns
self.assertTrue(
len(self.tm.mmGetTracePredictedInactiveColumns().data[15]) > 0)
self.assertTrue(
len(self.tm.mmGetTracePredictedInactiveColumns().data[35]) > 0)
def testH2(self):
"""Same as H1, but with cellsPerColumn == 4, and train multiple times.
It should make just the right number of predictions."""
self.init({"cellsPerColumn": 4})
numbers = self.sequenceMachine.generateNumbers(2, 20, (10, 15))
sequence = self.sequenceMachine.generateFromNumbers(numbers)
for _ in xrange(10):
self.feedTM(sequence)
self._testTM(sequence)
self.assertAllActiveWerePredicted()
# Without some kind of decay, expect predicted inactive columns at the
# end of the first shared sequence
predictedInactiveColumnsMetric = self.tm.mmGetMetricFromTrace(
self.tm.mmGetTracePredictedInactiveColumns())
self.assertTrue(predictedInactiveColumnsMetric.sum < 26)
# At the end of the second shared sequence, there should be no
# predicted but inactive columns
self.assertEqual(
len(self.tm.mmGetTracePredictedInactiveColumns().data[36]), 0)
def testH3(self):
"""Like H2, except the shared subsequence is in the beginning.
(e.g. "ABCDEF" and "ABCGHIJ") At the point where the shared subsequence
ends, all possible next patterns should be predicted. As soon as you see
the first unique pattern, the predictions should collapse to be a perfect
prediction."""
self.init({"cellsPerColumn": 4})
numbers = self.sequenceMachine.generateNumbers(2, 20, (0, 5))
sequence = self.sequenceMachine.generateFromNumbers(numbers)
self.feedTM(sequence)
self._testTM(sequence)
self.assertAllActiveWerePredicted()
predictedInactiveColumnsMetric = self.tm.mmGetMetricFromTrace(
self.tm.mmGetTracePredictedInactiveColumns())
self.assertTrue(predictedInactiveColumnsMetric.sum < 26 * 2)
# At the end of each shared sequence, there should be
# predicted but inactive columns
self.assertTrue(
len(self.tm.mmGetTracePredictedInactiveColumns().data[5]) > 0)
self.assertTrue(
len(self.tm.mmGetTracePredictedInactiveColumns().data[25]) > 0)
def testH4(self):
"""Shared patterns. Similar to H2 except that patterns are shared between
sequences. All sequences are different shufflings of the same set of N
patterns (there is no shared subsequence)."""
self.init({"cellsPerColumn": 4})
numbers = []
for _ in xrange(2):
numbers += self.sequenceMachine.generateNumbers(1, 20)
sequence = self.sequenceMachine.generateFromNumbers(numbers)
for _ in xrange(20):
self.feedTM(sequence)
self._testTM(sequence)
self.assertAllActiveWerePredicted()
predictedInactiveColumnsMetric = self.tm.mmGetMetricFromTrace(
self.tm.mmGetTracePredictedInactiveColumns())
self.assertTrue(predictedInactiveColumnsMetric.mean < 3)
def testH5(self):
"""Combination of H4) and H2).
Shared patterns in different sequences, with a shared subsequence."""
self.init({"cellsPerColumn": 4})
numbers = []
shared = self.sequenceMachine.generateNumbers(1, 5)[:-1]
for _ in xrange(2):
sublist = self.sequenceMachine.generateNumbers(1, 20)
sublist = [x for x in sublist if x not in xrange(5)]
numbers += sublist[0:10] + shared + sublist[10:]
sequence = self.sequenceMachine.generateFromNumbers(numbers)
for _ in xrange(20):
self.feedTM(sequence)
self._testTM(sequence)
self.assertAllActiveWerePredicted()
predictedInactiveColumnsMetric = self.tm.mmGetMetricFromTrace(
self.tm.mmGetTracePredictedInactiveColumns())
self.assertTrue(predictedInactiveColumnsMetric.mean < 3)
def testH9(self):
"""Sensitivity to small amounts of spatial noise during inference
(X = 0.05). Parameters the same as B11, and sequences like H2."""
self.init({"cellsPerColumn": 4,
"activationThreshold": 8})
numbers = self.sequenceMachine.generateNumbers(2, 20, (10, 15))
sequence = self.sequenceMachine.generateFromNumbers(numbers)
for _ in xrange(10):
self.feedTM(sequence)
sequence = self.sequenceMachine.addSpatialNoise(sequence, 0.05)
self._testTM(sequence)
unpredictedActiveColumnsMetric = self.tm.mmGetMetricFromTrace(
self.tm.mmGetTraceUnpredictedActiveColumns())
self.assertTrue(unpredictedActiveColumnsMetric.mean < 3)
def testH10(self):
"""Orphan Decay mechanism reduce predicted inactive cells (extra predictions).
Test feeds in noisy sequences (X = 0.05) to TM with and without orphan decay.
TM with orphan decay should has many fewer predicted inactive columns.
Parameters the same as B11, and sequences like H9."""
self.init({"cellsPerColumn": 4,
"activationThreshold": 8})
numbers = self.sequenceMachine.generateNumbers(2, 20, (10, 15))
sequence = self.sequenceMachine.generateFromNumbers(numbers)
sequenceNoisy = dict()
for i in xrange(10):
sequenceNoisy[i] = self.sequenceMachine.addSpatialNoise(sequence, 0.05)
self.feedTM(sequenceNoisy[i])
self._testTM(sequence)
predictedInactiveColumnsMetric = self.tm.mmGetMetricFromTrace(
self.tm.mmGetTracePredictedInactiveColumns())
predictedInactiveColumnsMean1 = predictedInactiveColumnsMetric.mean
self.init({"cellsPerColumn": 4,
"activationThreshold": 8,
"predictedSegmentDecrement": 0.004})
for _ in xrange(10):
self.feedTM(sequenceNoisy[i])
self._testTM(sequence)
predictedInactiveColumnsMetric = self.tm.mmGetMetricFromTrace(
self.tm.mmGetTracePredictedInactiveColumns())
predictedInactiveColumnsMean2 = predictedInactiveColumnsMetric.mean
self.assertGreater(predictedInactiveColumnsMean1, 0)
self.assertGreater(predictedInactiveColumnsMean1, predictedInactiveColumnsMean2)
# ==============================
# Overrides
# ==============================
def setUp(self):
super(ExtensiveTemporalMemoryTest, self).setUp()
print ("\n"
"======================================================\n"
"Test: {0} \n"
"{1}\n"
"======================================================\n"
).format(self.id(), self.shortDescription())
def feedTM(self, sequence, learn=True, num=1):
super(ExtensiveTemporalMemoryTest, self).feedTM(
sequence, learn=learn, num=num)
if self.VERBOSITY >= 2:
print self.tm.mmPrettyPrintTraces(
self.tm.mmGetDefaultTraces(verbosity=self.VERBOSITY-1))
print
if learn and self.VERBOSITY >= 3:
print self.tm.mmPrettyPrintConnections()
# ==============================
# Helper functions
# ==============================
def _testTM(self, sequence):
self.feedTM(sequence, learn=False)
print self.tm.mmPrettyPrintMetrics(self.tm.mmGetDefaultMetrics())
def assertAllActiveWerePredicted(self):
unpredictedActiveColumnsMetric = self.tm.mmGetMetricFromTrace(
self.tm.mmGetTraceUnpredictedActiveColumns())
predictedActiveColumnsMetric = self.tm.mmGetMetricFromTrace(
self.tm.mmGetTracePredictedActiveColumns())
self.assertEqual(unpredictedActiveColumnsMetric.sum, 0)
self.assertEqual(predictedActiveColumnsMetric.min, 21)
self.assertEqual(predictedActiveColumnsMetric.max, 25)
def assertAllInactiveWereUnpredicted(self):
predictedInactiveColumnsMetric = self.tm.mmGetMetricFromTrace(
self.tm.mmGetTracePredictedInactiveColumns())
self.assertEqual(predictedInactiveColumnsMetric.sum, 0)
def assertAllActiveWereUnpredicted(self):
unpredictedActiveColumnsMetric = self.tm.mmGetMetricFromTrace(
self.tm.mmGetTraceUnpredictedActiveColumns())
predictedActiveColumnsMetric = self.tm.mmGetMetricFromTrace(
self.tm.mmGetTracePredictedActiveColumns())
self.assertEqual(predictedActiveColumnsMetric.sum, 0)
self.assertEqual(unpredictedActiveColumnsMetric.min, 21)
self.assertEqual(unpredictedActiveColumnsMetric.max, 25)
def setUp(self):
self.patternMachine = PatternMachine(10000, 5, num=50)
class ExtendedTemporalMemoryTest(unittest.TestCase):
def setUp(self):
self.tm = ExtendedTemporalMemory(learnOnOneCell=False)
def testInitInvalidParams(self):
# Invalid columnDimensions
kwargs = {"columnDimensions": [], "cellsPerColumn": 32}
self.assertRaises(ValueError, ExtendedTemporalMemory, **kwargs)
# Invalid cellsPerColumn
kwargs = {"columnDimensions": [2048], "cellsPerColumn": 0}
self.assertRaises(ValueError, ExtendedTemporalMemory, **kwargs)
kwargs = {"columnDimensions": [2048], "cellsPerColumn": -10}
self.assertRaises(ValueError, ExtendedTemporalMemory, **kwargs)
def testlearnOnOneCellParam(self):
tm = self.tm
self.assertFalse(tm.learnOnOneCell)
tm = ExtendedTemporalMemory(learnOnOneCell=True)
self.assertTrue(tm.learnOnOneCell)
def testActivateCorrectlyPredictiveCells(self):
tm = self.tm
prevPredictiveCells = set([0, 237, 1026, 26337, 26339, 55536])
activeColumns = set([32, 47, 823])
prevMatchingCells = set()
(activeCells, winnerCells, predictedColumns, predictedInactiveCells) = tm.activateCorrectlyPredictiveCells(
prevPredictiveCells, prevMatchingCells, activeColumns
)
self.assertEqual(activeCells, set([1026, 26337, 26339]))
self.assertEqual(winnerCells, set([1026, 26337, 26339]))
self.assertEqual(predictedColumns, set([32, 823]))
self.assertEqual(predictedInactiveCells, set())
def testActivateCorrectlyPredictiveCellsEmpty(self):
tm = self.tm
# No previous predictive cells, no active columns
prevPredictiveCells = set()
activeColumns = set()
prevMatchingCells = set()
(activeCells, winnerCells, predictedColumns, predictedInactiveCells) = tm.activateCorrectlyPredictiveCells(
prevPredictiveCells, prevMatchingCells, activeColumns
)
self.assertEqual(activeCells, set())
self.assertEqual(winnerCells, set())
self.assertEqual(predictedColumns, set())
self.assertEqual(predictedInactiveCells, set())
# No previous predictive cells, with active columns
prevPredictiveCells = set()
activeColumns = set([32, 47, 823])
prevMatchingCells = set()
(activeCells, winnerCells, predictedColumns, predictedInactiveCells) = tm.activateCorrectlyPredictiveCells(
prevPredictiveCells, prevMatchingCells, activeColumns
)
self.assertEqual(activeCells, set())
self.assertEqual(winnerCells, set())
self.assertEqual(predictedColumns, set())
self.assertEqual(predictedInactiveCells, set())
# No active columns, with previously predictive cells
prevPredictiveCells = set([0, 237, 1026, 26337, 26339, 55536])
activeColumns = set()
prevMatchingCells = set()
(activeCells, winnerCells, predictedColumns, predictedInactiveCells) = tm.activateCorrectlyPredictiveCells(
prevPredictiveCells, prevMatchingCells, activeColumns
)
self.assertEqual(activeCells, set())
self.assertEqual(winnerCells, set())
self.assertEqual(predictedColumns, set())
self.assertEqual(predictedInactiveCells, set())
def testActivateCorrectlyPredictiveCellsOrphan(self):
tm = self.tm
tm.predictedSegmentDecrement = 0.001
prevPredictiveCells = set([])
activeColumns = set([32, 47, 823])
prevMatchingCells = set([32, 47])
(activeCells, winnerCells, predictedColumns, predictedInactiveCells) = tm.activateCorrectlyPredictiveCells(
prevPredictiveCells, prevMatchingCells, activeColumns
)
self.assertEqual(activeCells, set([]))
self.assertEqual(winnerCells, set([]))
self.assertEqual(predictedColumns, set([]))
self.assertEqual(predictedInactiveCells, set([32, 47]))
def testBurstColumns(self):
tm = ExtendedTemporalMemory(cellsPerColumn=4, connectedPermanence=0.50, minThreshold=1, seed=42)
connections = tm.connections
connections.createSegment(0)
connections.createSynapse(0, 23, 0.6)
connections.createSynapse(0, 37, 0.4)
connections.createSynapse(0, 477, 0.9)
connections.createSegment(0)
connections.createSynapse(1, 49, 0.9)
connections.createSynapse(1, 3, 0.8)
connections.createSegment(1)
connections.createSynapse(2, 733, 0.7)
connections.createSegment(108)
connections.createSynapse(3, 486, 0.9)
activeColumns = set([0, 1, 26])
predictedColumns = set([26])
prevActiveCells = set([23, 37, 49, 733])
prevWinnerCells = set([23, 37, 49, 733])
prevActiveApicalCells = set()
learnOnOneCell = False
chosenCellForColumn = {}
(activeCells, winnerCells, learningSegments, apicalLearningSegments, chosenCellForColumn) = tm.burstColumns(
activeColumns,
predictedColumns,
prevActiveCells,
prevActiveApicalCells,
prevWinnerCells,
learnOnOneCell,
chosenCellForColumn,
connections,
tm.apicalConnections,
)
self.assertEqual(activeCells, set([0, 1, 2, 3, 4, 5, 6, 7]))
randomWinner = 4 # 4 should be randomly chosen cell
self.assertEqual(winnerCells, set([0, randomWinner]))
self.assertEqual(learningSegments, set([0, 4])) # 4 is new segment created
# Check that new segment was added to winner cell (6) in column 1
self.assertEqual(connections.segmentsForCell(randomWinner), set([4]))
def testBurstColumnsEmpty(self):
tm = self.tm
activeColumns = set()
predictedColumns = set()
prevActiveCells = set()
prevWinnerCells = set()
connections = tm.connections
prevActiveApicalCells = set()
learnOnOneCell = False
chosenCellForColumn = {}
(activeCells, winnerCells, learningSegments, apicalLearningSegments, chosenCellForColumn) = tm.burstColumns(
activeColumns,
predictedColumns,
prevActiveCells,
prevActiveApicalCells,
prevWinnerCells,
learnOnOneCell,
chosenCellForColumn,
connections,
tm.apicalConnections,
)
self.assertEqual(activeCells, set())
self.assertEqual(winnerCells, set())
self.assertEqual(learningSegments, set())
self.assertEqual(apicalLearningSegments, set())
def testLearnOnSegments(self):
tm = ExtendedTemporalMemory(maxNewSynapseCount=2)
connections = tm.connections
connections.createSegment(0)
connections.createSynapse(0, 23, 0.6)
connections.createSynapse(0, 37, 0.4)
connections.createSynapse(0, 477, 0.9)
connections.createSegment(1)
connections.createSynapse(1, 733, 0.7)
connections.createSegment(8)
connections.createSynapse(2, 486, 0.9)
connections.createSegment(100)
prevActiveSegments = set([0, 2])
learningSegments = set([1, 3])
prevActiveCells = set([23, 37, 733])
winnerCells = set([0])
prevWinnerCells = set([10, 11, 12, 13, 14])
predictedInactiveCells = set()
prevMatchingSegments = set()
tm.learnOnSegments(
prevActiveSegments,
learningSegments,
prevActiveCells,
winnerCells,
prevWinnerCells,
connections,
predictedInactiveCells,
prevMatchingSegments,
)
# Check segment 0
synapseData = connections.dataForSynapse(0)
self.assertAlmostEqual(synapseData.permanence, 0.7)
synapseData = connections.dataForSynapse(1)
self.assertAlmostEqual(synapseData.permanence, 0.5)
synapseData = connections.dataForSynapse(2)
self.assertAlmostEqual(synapseData.permanence, 0.8)
# Check segment 1
synapseData = connections.dataForSynapse(3)
self.assertAlmostEqual(synapseData.permanence, 0.8)
self.assertEqual(len(connections.synapsesForSegment(1)), 2)
# Check segment 2
synapseData = connections.dataForSynapse(4)
self.assertAlmostEqual(synapseData.permanence, 0.9)
self.assertEqual(len(connections.synapsesForSegment(2)), 1)
# Check segment 3
self.assertEqual(len(connections.synapsesForSegment(3)), 2)
def testComputePredictiveCells(self):
tm = ExtendedTemporalMemory(activationThreshold=2, minThreshold=2, predictedSegmentDecrement=0.004)
connections = tm.connections
connections.createSegment(0)
connections.createSynapse(0, 23, 0.6)
connections.createSynapse(0, 37, 0.5)
connections.createSynapse(0, 477, 0.9)
connections.createSegment(1)
connections.createSynapse(1, 733, 0.7)
connections.createSynapse(1, 733, 0.4)
connections.createSegment(1)
connections.createSynapse(2, 974, 0.9)
connections.createSegment(8)
connections.createSynapse(3, 486, 0.9)
connections.createSegment(100)
activeCells = set([23, 37, 733, 974])
(activeSegments, predictiveCells, matchingSegments, matchingCells) = tm.computePredictiveCells(
activeCells, connections
)
self.assertEqual(activeSegments, set([0]))
self.assertEqual(predictiveCells, set([0]))
self.assertEqual(matchingSegments, set([0, 1]))
self.assertEqual(matchingCells, set([0, 1]))
def testBestMatchingCell(self):
tm = ExtendedTemporalMemory(connectedPermanence=0.50, minThreshold=1, seed=42)
connections = tm.connections
connections.createSegment(0)
connections.createSynapse(0, 23, 0.6)
connections.createSynapse(0, 37, 0.4)
connections.createSynapse(0, 477, 0.9)
connections.createSegment(0)
connections.createSynapse(1, 49, 0.9)
connections.createSynapse(1, 3, 0.8)
connections.createSegment(1)
connections.createSynapse(2, 733, 0.7)
connections.createSegment(108)
connections.createSynapse(3, 486, 0.9)
activeCells = set([23, 37, 49, 733])
activeApicalCells = set()
self.assertEqual(
tm.bestMatchingCell(
tm.cellsForColumn(0), activeCells, activeApicalCells, connections, tm.apicalConnections
),
(0, 0, None),
)
self.assertEqual(
tm.bestMatchingCell(
tm.cellsForColumn(3), activeCells, activeApicalCells, connections, tm.apicalConnections
),
(103, None, None),
) # Random cell from column
self.assertEqual(
tm.bestMatchingCell(
tm.cellsForColumn(999), activeCells, activeApicalCells, connections, tm.apicalConnections
),
(31979, None, None),
) # Random cell from column
def testBestMatchingCellFewestSegments(self):
tm = ExtendedTemporalMemory(
columnDimensions=[2], cellsPerColumn=2, connectedPermanence=0.50, minThreshold=1, seed=42
)
connections = tm.connections
connections.createSegment(0)
connections.createSynapse(0, 3, 0.3)
activeSynapsesForSegment = set([])
activeApicalCells = set()
for _ in range(100):
# Never pick cell 0, always pick cell 1
(cell, _, _) = tm.bestMatchingCell(
tm.cellsForColumn(0), activeSynapsesForSegment, activeApicalCells, connections, tm.apicalConnections
)
self.assertEqual(cell, 1)
def testBestMatchingSegment(self):
tm = ExtendedTemporalMemory(connectedPermanence=0.50, minThreshold=1)
connections = tm.connections
connections.createSegment(0)
connections.createSynapse(0, 23, 0.6)
connections.createSynapse(0, 37, 0.4)
connections.createSynapse(0, 477, 0.9)
connections.createSegment(0)
connections.createSynapse(1, 49, 0.9)
connections.createSynapse(1, 3, 0.8)
connections.createSegment(1)
connections.createSynapse(2, 733, 0.7)
connections.createSegment(8)
connections.createSynapse(3, 486, 0.9)
activeCells = set([23, 37, 49, 733])
self.assertEqual(tm.bestMatchingSegment(0, activeCells, connections), (0, 2))
self.assertEqual(tm.bestMatchingSegment(1, activeCells, connections), (2, 1))
self.assertEqual(tm.bestMatchingSegment(8, activeCells, connections), (None, None))
self.assertEqual(tm.bestMatchingSegment(100, activeCells, connections), (None, None))
def testLeastUsedCell(self):
tm = ExtendedTemporalMemory(columnDimensions=[2], cellsPerColumn=2, seed=42)
connections = tm.connections
connections.createSegment(0)
connections.createSynapse(0, 3, 0.3)
for _ in range(100):
# Never pick cell 0, always pick cell 1
self.assertEqual(tm.leastUsedCell(tm.cellsForColumn(0), connections), 1)
def testAdaptSegment(self):
tm = self.tm
connections = tm.connections
connections.createSegment(0)
connections.createSynapse(0, 23, 0.6)
connections.createSynapse(0, 37, 0.4)
connections.createSynapse(0, 477, 0.9)
tm.adaptSegment(0, set([0, 1]), connections, tm.permanenceIncrement, tm.permanenceDecrement)
synapseData = connections.dataForSynapse(0)
self.assertAlmostEqual(synapseData.permanence, 0.7)
synapseData = connections.dataForSynapse(1)
self.assertAlmostEqual(synapseData.permanence, 0.5)
synapseData = connections.dataForSynapse(2)
self.assertAlmostEqual(synapseData.permanence, 0.8)
def testAdaptSegmentToMax(self):
tm = self.tm
connections = tm.connections
connections.createSegment(0)
connections.createSynapse(0, 23, 0.9)
tm.adaptSegment(0, set([0]), connections, tm.permanenceIncrement, tm.permanenceDecrement)
synapseData = connections.dataForSynapse(0)
self.assertAlmostEqual(synapseData.permanence, 1.0)
# Now permanence should be at max
tm.adaptSegment(0, set([0]), connections, tm.permanenceIncrement, tm.permanenceDecrement)
synapseData = connections.dataForSynapse(0)
self.assertAlmostEqual(synapseData.permanence, 1.0)
def testAdaptSegmentToMin(self):
tm = self.tm
connections = tm.connections
connections.createSegment(0)
connections.createSynapse(0, 23, 0.1)
tm.adaptSegment(0, set(), connections, tm.permanenceIncrement, tm.permanenceDecrement)
synapses = connections.synapsesForSegment(0)
self.assertFalse(0 in synapses)
def testPickCellsToLearnOn(self):
tm = ExtendedTemporalMemory(seed=42)
connections = tm.connections
connections.createSegment(0)
winnerCells = set([4, 47, 58, 93])
self.assertEqual(tm.pickCellsToLearnOn(2, 0, winnerCells, connections), set([4, 93])) # randomly picked
self.assertEqual(tm.pickCellsToLearnOn(100, 0, winnerCells, connections), set([4, 47, 58, 93]))
self.assertEqual(tm.pickCellsToLearnOn(0, 0, winnerCells, connections), set())
def testPickCellsToLearnOnAvoidDuplicates(self):
tm = ExtendedTemporalMemory(seed=42)
connections = tm.connections
connections.createSegment(0)
connections.createSynapse(0, 23, 0.6)
winnerCells = set([23])
# Ensure that no additional (duplicate) cells were picked
self.assertEqual(tm.pickCellsToLearnOn(2, 0, winnerCells, connections), set())
def testColumnForCell1D(self):
tm = ExtendedTemporalMemory(columnDimensions=[2048], cellsPerColumn=5)
self.assertEqual(tm.columnForCell(0), 0)
self.assertEqual(tm.columnForCell(4), 0)
self.assertEqual(tm.columnForCell(5), 1)
self.assertEqual(tm.columnForCell(10239), 2047)
def testColumnForCell2D(self):
tm = ExtendedTemporalMemory(columnDimensions=[64, 64], cellsPerColumn=4)
self.assertEqual(tm.columnForCell(0), 0)
self.assertEqual(tm.columnForCell(3), 0)
self.assertEqual(tm.columnForCell(4), 1)
self.assertEqual(tm.columnForCell(16383), 4095)
def testColumnForCellInvalidCell(self):
tm = ExtendedTemporalMemory(columnDimensions=[64, 64], cellsPerColumn=4)
try:
tm.columnForCell(16383)
except IndexError:
self.fail("IndexError raised unexpectedly")
args = [16384]
self.assertRaises(IndexError, tm.columnForCell, *args)
args = [-1]
self.assertRaises(IndexError, tm.columnForCell, *args)
def testCellsForColumn1D(self):
tm = ExtendedTemporalMemory(columnDimensions=[2048], cellsPerColumn=5)
expectedCells = set([5, 6, 7, 8, 9])
self.assertEqual(tm.cellsForColumn(1), expectedCells)
def testCellsForColumn2D(self):
tm = ExtendedTemporalMemory(columnDimensions=[64, 64], cellsPerColumn=4)
expectedCells = set([256, 257, 258, 259])
self.assertEqual(tm.cellsForColumn(64), expectedCells)
def testCellsForColumnInvalidColumn(self):
tm = ExtendedTemporalMemory(columnDimensions=[64, 64], cellsPerColumn=4)
try:
tm.cellsForColumn(4095)
except IndexError:
self.fail("IndexError raised unexpectedly")
args = [4096]
self.assertRaises(IndexError, tm.cellsForColumn, *args)
args = [-1]
self.assertRaises(IndexError, tm.cellsForColumn, *args)
def testNumberOfColumns(self):
tm = ExtendedTemporalMemory(columnDimensions=[64, 64], cellsPerColumn=32)
self.assertEqual(tm.numberOfColumns(), 64 * 64)
def testNumberOfCells(self):
tm = ExtendedTemporalMemory(columnDimensions=[64, 64], cellsPerColumn=32)
self.assertEqual(tm.numberOfCells(), 64 * 64 * 32)
def testMapCellsToColumns(self):
tm = ExtendedTemporalMemory(columnDimensions=[100], cellsPerColumn=4)
columnsForCells = tm.mapCellsToColumns(set([0, 1, 2, 5, 399]))
self.assertEqual(columnsForCells[0], set([0, 1, 2]))
self.assertEqual(columnsForCells[1], set([5]))
self.assertEqual(columnsForCells[99], set([399]))
def testCalculatePredictiveCells(self):
tm = ExtendedTemporalMemory(columnDimensions=[4], cellsPerColumn=5)
predictiveDistalCells = set([2, 3, 5, 8, 10, 12, 13, 14])
predictiveApicalCells = set([1, 5, 7, 11, 14, 15, 17])
self.assertEqual(tm.calculatePredictiveCells(predictiveDistalCells, predictiveApicalCells), set([2, 3, 5, 14]))
def testCompute(self):
tm = ExtendedTemporalMemory(
columnDimensions=[4],
cellsPerColumn=10,
learnOnOneCell=False,
initialPermanence=0.2,
connectedPermanence=0.7,
activationThreshold=1,
)
seg1 = tm.connections.createSegment(0)
seg2 = tm.connections.createSegment(20)
seg3 = tm.connections.createSegment(25)
try:
tm.connections.createSynapse(seg1, 15, 0.9)
tm.connections.createSynapse(seg2, 35, 0.9)
tm.connections.createSynapse(seg2, 45, 0.9) # external cell
tm.connections.createSynapse(seg3, 35, 0.9)
tm.connections.createSynapse(seg3, 50, 0.9) # external cell
except IndexError:
self.fail("IndexError raised unexpectedly for distal segments")
aSeg1 = tm.apicalConnections.createSegment(1)
aSeg2 = tm.apicalConnections.createSegment(25)
try:
tm.apicalConnections.createSynapse(aSeg1, 3, 0.9)
tm.apicalConnections.createSynapse(aSeg2, 1, 0.9)
except IndexError:
self.fail("IndexError raised unexpectedly for apical segments")
activeColumns = set([1, 3])
activeExternalCells = set([5, 10, 15])
activeApicalCells = set([1, 2, 3, 4])
tm.compute(
activeColumns, activeExternalCells=activeExternalCells, activeApicalCells=activeApicalCells, learn=False
)
activeColumns = set([0, 2])
tm.compute(activeColumns, activeExternalCells=set(), activeApicalCells=set())
self.assertEqual(tm.activeCells, set([0, 20, 25]))
def testLearning(self):
tm = ExtendedTemporalMemory(
columnDimensions=[4],
cellsPerColumn=10,
learnOnOneCell=False,
initialPermanence=0.5,
connectedPermanence=0.6,
activationThreshold=1,
minThreshold=1,
maxNewSynapseCount=2,
permanenceDecrement=0.05,
permanenceIncrement=0.2,
)
seg1 = tm.connections.createSegment(0)
seg2 = tm.connections.createSegment(10)
seg3 = tm.connections.createSegment(20)
seg4 = tm.connections.createSegment(30)
try:
tm.connections.createSynapse(seg1, 10, 0.9)
tm.connections.createSynapse(seg2, 20, 0.9)
tm.connections.createSynapse(seg3, 30, 0.9)
tm.connections.createSynapse(seg3, 41, 0.9)
tm.connections.createSynapse(seg3, 25, 0.9)
tm.connections.createSynapse(seg4, 0, 0.9)
except IndexError:
self.fail("IndexError raised unexpectedly for distal segments")
aSeg1 = tm.apicalConnections.createSegment(0)
aSeg2 = tm.apicalConnections.createSegment(20)
try:
tm.apicalConnections.createSynapse(aSeg1, 42, 0.8)
tm.apicalConnections.createSynapse(aSeg2, 43, 0.8)
except IndexError:
self.fail("IndexError raised unexpectedly for apical segments")
activeColumns = set([1, 3])
activeExternalCells = set([1]) # will be re-indexed to 41
activeApicalCells = set([2, 3]) # will be re-indexed to 42, 43
tm.compute(
activeColumns, activeExternalCells=activeExternalCells, activeApicalCells=activeApicalCells, learn=False
)
activeColumns = set([0, 2])
tm.compute(activeColumns, activeExternalCells=None, activeApicalCells=None, learn=True)
self.assertEqual(tm.activeCells, set([0, 20]))
# distal learning
synapse = list(tm.connections.synapsesForSegment(seg1))[0]
self.assertEqual(tm.connections.dataForSynapse(synapse).permanence, 1.0)
synapse = list(tm.connections.synapsesForSegment(seg2))[0]
self.assertEqual(tm.connections.dataForSynapse(synapse).permanence, 0.9)
synapse = list(tm.connections.synapsesForSegment(seg3))[0]
self.assertEqual(tm.connections.dataForSynapse(synapse).permanence, 1.0)
synapse = list(tm.connections.synapsesForSegment(seg3))[1]
self.assertEqual(tm.connections.dataForSynapse(synapse).permanence, 1.0)
synapse = list(tm.connections.synapsesForSegment(seg3))[2]
self.assertEqual(tm.connections.dataForSynapse(synapse).permanence, 0.85)
synapse = list(tm.connections.synapsesForSegment(seg4))[0]
self.assertEqual(tm.connections.dataForSynapse(synapse).permanence, 0.9)
# apical learning
synapse = list(tm.apicalConnections.synapsesForSegment(aSeg1))[0]
self.assertEqual(tm.apicalConnections.dataForSynapse(synapse).permanence, 1.0)
synapse = list(tm.apicalConnections.synapsesForSegment(aSeg2))[0]
self.assertEqual(tm.apicalConnections.dataForSynapse(synapse).permanence, 1.0)
@unittest.skipUnless(capnp is not None, "No serialization available for ETM")
def testWriteRead(self):
tm1 = ExtendedTemporalMemory(
columnDimensions=[100],
cellsPerColumn=4,
activationThreshold=7,
initialPermanence=0.37,
connectedPermanence=0.58,
minThreshold=4,
maxNewSynapseCount=18,
permanenceIncrement=0.23,
permanenceDecrement=0.08,
seed=91,
)
# Run some data through before serializing
self.patternMachine = PatternMachine(100, 4)
self.sequenceMachine = SequenceMachine(self.patternMachine)
sequence = self.sequenceMachine.generateFromNumbers(range(5))
for _ in range(3):
for pattern in sequence:
tm1.compute(pattern)
proto1 = TemporalMemoryProto_capnp.TemporalMemoryProto.new_message()
tm1.write(proto1)
# Write the proto to a temp file and read it back into a new proto
with tempfile.TemporaryFile() as f:
proto1.write(f)
f.seek(0)
proto2 = TemporalMemoryProto_capnp.TemporalMemoryProto.read(f)
# Load the deserialized proto
tm2 = ExtendedTemporalMemory.read(proto2)
# Check that the two temporal memory objects have the same attributes
self.assertEqual(tm1, tm2)
# Run a couple records through after deserializing and check results match
tm1.compute(self.patternMachine.get(0))
tm2.compute(self.patternMachine.get(0))
self.assertEqual(set(tm1.getActiveCells()), set(tm2.getActiveCells()))
self.assertEqual(set(tm1.getPredictiveCells()), set(tm2.getPredictiveCells()))
self.assertEqual(set(tm1.getWinnerCells()), set(tm2.getWinnerCells()))
self.assertEqual(tm1.connections, tm2.connections)
tm1.compute(self.patternMachine.get(3))
tm2.compute(self.patternMachine.get(3))
self.assertEqual(set(tm1.getActiveCells()), set(tm2.getActiveCells()))
self.assertEqual(set(tm1.getPredictiveCells()), set(tm2.getPredictiveCells()))
self.assertEqual(set(tm1.getWinnerCells()), set(tm2.getWinnerCells()))
self.assertEqual(tm1.connections, tm2.connections)
class ExtensiveColumnPoolerTest(unittest.TestCase):
"""
Algorithmic tests for the ColumnPooler region.
Each test actually tests multiple aspects of the algorithm. For more
atomic tests refer to column_pooler_unit_test.
The notation for objects is the following:
object{patternA, patternB, ...}
In these tests, the proximally-fed SDR's are simulated as unique (location,
feature) pairs regardless of actual locations and features, unless stated
otherwise.
"""
inputWidth = 2048 * 8
numInputActiveBits = int(0.02 * inputWidth)
outputWidth = 2048
numOutputActiveBits = 40
seed = 42
def testNewInputs(self):
"""
Checks that the behavior is correct when facing unseed inputs.
"""
self.init()
# feed the first input, a random SDR should be generated
initialPattern = self.generateObject(1)
self.learn(initialPattern, numRepetitions=1, newObject=True)
representation = self._getActiveRepresentation()
self.assertEqual(
len(representation),
self.numOutputActiveBits,
"The generated representation is incorrect"
)
# feed a new input for the same object, the previous SDR should persist
newPattern = self.generateObject(1)
self.learn(newPattern, numRepetitions=1, newObject=False)
newRepresentation = self._getActiveRepresentation()
self.assertNotEqual(initialPattern, newPattern)
self.assertEqual(
newRepresentation,
representation,
"The SDR did not persist when learning the same object"
)
# without sensory input, the SDR should persist as well
emptyPattern = [set()]
self.learn(emptyPattern, numRepetitions=1, newObject=False)
newRepresentation = self._getActiveRepresentation()
self.assertEqual(
newRepresentation,
representation,
"The SDR did not persist after an empty input."
)
def testLearnSinglePattern(self):
"""
A single pattern is learnt for a single object.
Objects: A{X, Y}
"""
self.init()
object = self.generateObject(1)
self.learn(object, numRepetitions=1, newObject=True)
# check that the active representation is sparse
representation = self._getActiveRepresentation()
self.assertEqual(
len(representation),
self.numOutputActiveBits,
"The generated representation is incorrect"
)
# check that the pattern was correctly learnt
self.infer(feedforwardPattern=object[0])
self.assertEqual(
self._getActiveRepresentation(),
representation,
"The pooled representation is not stable"
)
# present new pattern for same object
# it should be mapped to the same representation
newPattern = [self.generatePattern()]
self.learn(newPattern, numRepetitions=1, newObject=False)
# check that the active representation is sparse
newRepresentation = self._getActiveRepresentation()
self.assertEqual(
newRepresentation,
representation,
"The new pattern did not map to the same object representation"
)
# check that the pattern was correctly learnt and is stable
self.infer(feedforwardPattern=object[0])
self.assertEqual(
self._getActiveRepresentation(),
representation,
"The pooled representation is not stable"
)
def testLearnSingleObject(self):
"""
Many patterns are learnt for a single object.
Objects: A{P, Q, R, S, T}
"""
self.init()
object = self.generateObject(numPatterns=5)
self.learn(object, numRepetitions=1, randomOrder=True, newObject=True)
representation = self._getActiveRepresentation()
# check that all patterns map to the same object
for pattern in object:
self.infer(feedforwardPattern=pattern)
self.assertEqual(
self._getActiveRepresentation(),
representation,
"The pooled representation is not stable"
)
# if activity stops, check that the representation persists
self.infer(feedforwardPattern=set())
self.assertEqual(
self._getActiveRepresentation(),
representation,
"The pooled representation did not persist"
)
def testLearnTwoObjectNoCommonPattern(self):
"""
Same test as before, using two objects, without common pattern.
Objects: A{P, Q, R, S,T} B{V, W, X, Y, Z}
"""
self.init()
objectA = self.generateObject(numPatterns=5)
self.learn(objectA, numRepetitions=1, randomOrder=True, newObject=True)
representationA = self._getActiveRepresentation()
objectB = self.generateObject(numPatterns=5)
self.learn(objectB, numRepetitions=1, randomOrder=True, newObject=True)
representationB = self._getActiveRepresentation()
self.assertNotEqual(representationA, representationB)
# check that all patterns map to the same object
for pattern in objectA:
self.infer(feedforwardPattern=pattern)
self.assertEqual(
self._getActiveRepresentation(),
representationA,
"The pooled representation for the first object is not stable"
)
# check that all patterns map to the same object
for pattern in objectB:
self.infer(feedforwardPattern=pattern)
self.assertEqual(
self._getActiveRepresentation(),
representationB,
"The pooled representation for the second object is not stable"
)
# feed union of patterns in object A
pattern = objectA[0] | objectA[1]
self.infer(feedforwardPattern=pattern)
self.assertEqual(
self._getActiveRepresentation(),
representationA,
"The active representation is incorrect"
)
# feed unions of patterns in objects A and B
pattern = objectA[0] | objectB[0]
self.infer(feedforwardPattern=pattern)
self.assertEqual(
self._getActiveRepresentation(),
representationA | representationB,
"The active representation is incorrect"
)
def testLearnTwoObjectsOneCommonPattern(self):
"""
Same test as before, except the two objects share a pattern
Objects: A{P, Q, R, S,T} B{P, W, X, Y, Z}
"""
self.init()
objectA = self.generateObject(numPatterns=5)
self.learn(objectA, numRepetitions=1, randomOrder=True, newObject=True)
representationA = self._getActiveRepresentation()
objectB = self.generateObject(numPatterns=5)
objectB[0] = objectA[0]
self.learn(objectB, numRepetitions=1, randomOrder=True, newObject=True)
representationB = self._getActiveRepresentation()
self.assertNotEqual(representationA, representationB)
# very small overlap
self.assertLessEqual(len(representationA & representationB), 3)
# check that all patterns except the common one map to the same object
for pattern in objectA[1:]:
self.infer(feedforwardPattern=pattern)
self.assertEqual(
self._getActiveRepresentation(),
representationA,
"The pooled representation for the first object is not stable"
)
# check that all patterns except the common one map to the same object
for pattern in objectB[1:]:
self.infer(feedforwardPattern=pattern)
self.assertEqual(
self._getActiveRepresentation(),
representationB,
"The pooled representation for the second object is not stable"
)
# feed shared pattern
pattern = objectA[0]
self.infer(feedforwardPattern=pattern)
self.assertEqual(
self._getActiveRepresentation(),
representationA | representationB,
"The active representation is incorrect"
)
# feed union of patterns in object A
pattern = objectA[1] | objectA[2]
self.infer(feedforwardPattern=pattern)
self.assertEqual(
self._getActiveRepresentation(),
representationA,
"The active representation is incorrect"
)
# feed unions of patterns in objects A and B
pattern = objectA[1] | objectB[1]
self.infer(feedforwardPattern=pattern)
self.assertEqual(
self._getActiveRepresentation(),
representationA | representationB,
"The active representation is incorrect"
)
def testLearnThreeObjectsOneCommonPattern(self):
"""
Same test as before, with three objects
Objects: A{P, Q, R, S,T} B{P, W, X, Y, Z} C{W, H, I, K, L}
"""
self.init()
objectA = self.generateObject(numPatterns=5)
self.learn(objectA, numRepetitions=1, randomOrder=True, newObject=True)
representationA = self._getActiveRepresentation()
objectB = self.generateObject(numPatterns=5)
objectB[0] = objectA[0]
self.learn(objectB, numRepetitions=1, randomOrder=True, newObject=True)
representationB = self._getActiveRepresentation()
objectC = self.generateObject(numPatterns=5)
objectC[0] = objectB[1]
self.learn(objectC, numRepetitions=1, randomOrder=True, newObject=True)
representationC = self._getActiveRepresentation()
self.assertNotEquals(representationA, representationB, representationC)
# very small overlap
self.assertLessEqual(len(representationA & representationB), 3)
self.assertLessEqual(len(representationB & representationC), 3)
self.assertLessEqual(len(representationA & representationC), 3)
# check that all patterns except the common one map to the same object
for pattern in objectA[1:]:
self.infer(feedforwardPattern=pattern)
self.assertEqual(
self._getActiveRepresentation(),
representationA,
"The pooled representation for the first object is not stable"
)
# check that all patterns except the common one map to the same object
for pattern in objectB[2:]:
self.infer(feedforwardPattern=pattern)
self.assertEqual(
self._getActiveRepresentation(),
representationB,
"The pooled representation for the second object is not stable"
)
# check that all patterns except the common one map to the same object
for pattern in objectC[1:]:
self.infer(feedforwardPattern=pattern)
self.assertEqual(
self._getActiveRepresentation(),
representationC,
"The pooled representation for the third object is not stable"
)
# feed shared pattern between A and B
pattern = objectA[0]
self.infer(feedforwardPattern=pattern)
self.assertEqual(
self._getActiveRepresentation(),
representationA | representationB,
"The active representation is incorrect"
)
# feed shared pattern between B and C
pattern = objectB[1]
self.infer(feedforwardPattern=pattern)
self.assertEqual(
self._getActiveRepresentation(),
representationB | representationC,
"The active representation is incorrect"
)
# feed union of patterns in object A
pattern = objectA[1] | objectA[2]
self.infer(feedforwardPattern=pattern)
self.assertEqual(
self._getActiveRepresentation(),
representationA,
"The active representation is incorrect"
)
# feed unions of patterns to activate all objects
pattern = objectA[1] | objectB[1]
self.infer(feedforwardPattern=pattern)
self.assertEqual(
self._getActiveRepresentation(),
representationA | representationB | representationC,
"The active representation is incorrect"
)
def testLearnThreeObjectsOneCommonPatternSpatialNoise(self):
"""
Same test as before, with three objects
Objects: A{P, Q, R, S,T} B{P, W, X, Y, Z} C{W, H, I, K, L}
"""
self.init()
objectA = self.generateObject(numPatterns=5)
self.learn(objectA, numRepetitions=1, randomOrder=True, newObject=True)
representationA = self._getActiveRepresentation()
objectB = self.generateObject(numPatterns=5)
objectB[0] = objectA[0]
self.learn(objectB, numRepetitions=1, randomOrder=True, newObject=True)
representationB = self._getActiveRepresentation()
objectC = self.generateObject(numPatterns=5)
objectC[0] = objectB[1]
self.learn(objectC, numRepetitions=1, randomOrder=True, newObject=True)
representationC = self._getActiveRepresentation()
self.assertNotEquals(representationA, representationB, representationC)
# very small overlap
self.assertLessEqual(len(representationA & representationB), 3)
self.assertLessEqual(len(representationB & representationC), 3)
self.assertLessEqual(len(representationA & representationC), 3)
# check that all patterns except the common one map to the same object
for pattern in objectA[1:]:
noisyPattern = self.proximalPatternMachine.addNoise(pattern, 0.05)
self.infer(feedforwardPattern=noisyPattern)
self.assertEqual(
self._getActiveRepresentation(),
representationA,
"The pooled representation for the first object is not stable"
)
# check that all patterns except the common one map to the same object
for pattern in objectB[2:]:
noisyPattern = self.proximalPatternMachine.addNoise(pattern, 0.05)
self.infer(feedforwardPattern=noisyPattern)
self.assertEqual(
self._getActiveRepresentation(),
representationB,
"The pooled representation for the second object is not stable"
)
# check that all patterns except the common one map to the same object
for pattern in objectC[1:]:
noisyPattern = self.proximalPatternMachine.addNoise(pattern, 0.05)
self.infer(feedforwardPattern=noisyPattern)
self.assertEqual(
self._getActiveRepresentation(),
representationC,
"The pooled representation for the third object is not stable"
)
# feed shared pattern between A and B
pattern = objectA[0]
noisyPattern = self.proximalPatternMachine.addNoise(pattern, 0.05)
self.infer(feedforwardPattern=noisyPattern)
self.assertEqual(
self._getActiveRepresentation(),
representationA | representationB,
"The active representation is incorrect"
)
# feed shared pattern between B and C
pattern = objectB[1]
noisyPattern = self.proximalPatternMachine.addNoise(pattern, 0.05)
self.infer(feedforwardPattern=noisyPattern)
self.assertEqual(
self._getActiveRepresentation(),
representationB | representationC,
"The active representation is incorrect"
)
# feed union of patterns in object A
pattern = objectA[1] | objectA[2]
noisyPattern = self.proximalPatternMachine.addNoise(pattern, 0.05)
self.infer(feedforwardPattern=noisyPattern)
self.assertEqual(
self._getActiveRepresentation(),
representationA,
"The active representation is incorrect"
)
# feed unions of patterns to activate all objects
pattern = objectA[1] | objectB[1]
noisyPattern = self.proximalPatternMachine.addNoise(pattern, 0.05)
self.infer(feedforwardPattern=noisyPattern)
self.assertEqual(
self._getActiveRepresentation(),
representationA | representationB | representationC,
"The active representation is incorrect"
)
def setUp(self):
"""
Sets up the test.
"""
self.pooler = None
self.proximalPatternMachine = PatternMachine(
n=self.inputWidth,
w=self.numOutputActiveBits,
num=200,
seed=self.seed
)
self.patternId = 0
np.random.seed(self.seed)
# Wrappers around ColumnPooler API
def learn(self,
feedforwardPatterns,
lateralPatterns=None,
numRepetitions=1,
randomOrder=True,
newObject=True):
"""
Parameters:
----------------------------
Learns a single object, with the provided patterns.
@param feedforwardPatterns (list(set))
List of proximal input patterns
@param lateralPatterns (list(list(set)))
List of distal input patterns, or None. If no lateral input is
used. The outer list is expected to have the same length as
feedforwardPatterns, whereas each inner list's length is the
number of cortical columns which are distally connected to the
pooler.
@param numRepetitions (int)
Number of times the patterns will be fed
@param randomOrder (bool)
If true, the order of patterns will be shuffled at each
repetition
"""
if newObject:
self.pooler.mmClearHistory()
self.pooler.reset()
# set-up
indices = range(len(feedforwardPatterns))
if lateralPatterns is None:
lateralPatterns = [None] * len(feedforwardPatterns)
for _ in xrange(numRepetitions):
if randomOrder:
np.random.shuffle(indices)
for idx in indices:
self.pooler.compute(feedforwardPatterns[idx],
activeExternalCells=lateralPatterns[idx],
learn=True)
def infer(self,
feedforwardPattern,
lateralPatterns=None,
printMetrics=False):
"""
Feeds a single pattern to the column pooler (as well as an eventual lateral
pattern).
Parameters:
----------------------------
@param feedforwardPattern (set)
Input proximal pattern to the pooler
@param lateralPatterns (list(set))
Input dislal patterns to the pooler (one for each neighboring CC's)
@param printMetrics (bool)
If true, will print cell metrics
"""
self.pooler.compute(feedforwardPattern,
activeExternalCells=lateralPatterns,
learn=False)
if printMetrics:
print self.pooler.mmPrettyPrintMetrics(
self.pooler.mmGetDefaultMetrics()
)
# Helper functions
def generatePattern(self):
"""
Returns a random proximal input pattern.
"""
pattern = self.proximalPatternMachine.get(self.patternId)
self.patternId += 1
return pattern
def generateObject(self, numPatterns):
"""
Creates a list of patterns, for a given object.
"""
return [self.generatePattern() for _ in xrange(numPatterns)]
def init(self, overrides=None):
"""
Creates the column pooler with specified parameter overrides.
Except for the specified overrides and problem-specific parameters, used
parameters are implementation defaults.
"""
params = {
"inputWidth": self.inputWidth,
"numActivecolumnsPerInhArea": self.numOutputActiveBits,
"columnDimensions": (self.outputWidth,),
"seed": self.seed,
"learnOnOneCell": False
}
if overrides is None:
overrides = {}
params.update(overrides)
self.pooler = MonitoredColumnPooler(**params)
def _getActiveRepresentation(self):
"""
Retrieves the current active representation in the pooler.
"""
if self.pooler is None:
raise ValueError("No pooler has been instantiated")
return set(self.pooler.getActiveCells())
def getPatternMachine(self):
return PatternMachine(100, range(21, 26), num=300)
class TemporalMemoryTest(unittest.TestCase):
def setUp(self):
self.tm = TemporalMemory()
def testInitInvalidParams(self):
# Invalid columnDimensions
kwargs = {"columnDimensions": [], "cellsPerColumn": 32}
self.assertRaises(ValueError, TemporalMemory, **kwargs)
# Invalid cellsPerColumn
kwargs = {"columnDimensions": [2048], "cellsPerColumn": 0}
self.assertRaises(ValueError, TemporalMemory, **kwargs)
kwargs = {"columnDimensions": [2048], "cellsPerColumn": -10}
self.assertRaises(ValueError, TemporalMemory, **kwargs)
def testActivateCorrectlyPredictiveCells(self):
tm = self.tm
prevPredictiveCells = set([0, 237, 1026, 26337, 26339, 55536])
activeColumns = set([32, 47, 823])
prevMatchingCells = set()
(activeCells,
winnerCells,
predictedColumns,
predictedInactiveCells) = tm.activateCorrectlyPredictiveCells(prevPredictiveCells,
prevMatchingCells,
activeColumns)
self.assertEqual(activeCells, set([1026, 26337, 26339]))
self.assertEqual(winnerCells, set([1026, 26337, 26339]))
self.assertEqual(predictedColumns, set([32, 823]))
self.assertEqual(predictedInactiveCells, set())
def testActivateCorrectlyPredictiveCellsEmpty(self):
tm = self.tm
# No previous predictive cells, no active columns
prevPredictiveCells = set()
activeColumns = set()
prevMatchingCells = set()
(activeCells,
winnerCells,
predictedColumns,
predictedInactiveCells) = tm.activateCorrectlyPredictiveCells(prevPredictiveCells,
prevMatchingCells,
activeColumns)
self.assertEqual(activeCells, set())
self.assertEqual(winnerCells, set())
self.assertEqual(predictedColumns, set())
self.assertEqual(predictedInactiveCells, set())
# No previous predictive cells, with active columns
prevPredictiveCells = set()
activeColumns = set([32, 47, 823])
prevMatchingCells = set()
(activeCells,
winnerCells,
predictedColumns,
predictedInactiveCells) = tm.activateCorrectlyPredictiveCells(prevPredictiveCells,
prevMatchingCells,
activeColumns)
self.assertEqual(activeCells, set())
self.assertEqual(winnerCells, set())
self.assertEqual(predictedColumns, set())
self.assertEqual(predictedInactiveCells, set())
# No active columns, with previously predictive cells
prevPredictiveCells = set([0, 237, 1026, 26337, 26339, 55536])
activeColumns = set()
prevMatchingCells = set()
(activeCells,
winnerCells,
predictedColumns,
predictedInactiveCells) = tm.activateCorrectlyPredictiveCells(prevPredictiveCells,
prevMatchingCells,
activeColumns)
self.assertEqual(activeCells, set())
self.assertEqual(winnerCells, set())
self.assertEqual(predictedColumns, set())
self.assertEqual(predictedInactiveCells, set())
def testActivateCorrectlyPredictiveCellsOrphan(self):
tm = self.tm
prevPredictiveCells = set([])
activeColumns = set([32, 47, 823])
prevMatchingCells = set([32, 47])
(activeCells,
winnerCells,
predictedColumns,
predictedInactiveCells) = tm.activateCorrectlyPredictiveCells(prevPredictiveCells,
prevMatchingCells,
activeColumns)
self.assertEqual(activeCells, set([]))
self.assertEqual(winnerCells, set([]))
self.assertEqual(predictedColumns, set([]))
self.assertEqual(predictedInactiveCells, set([32,47]))
def testBurstColumns(self):
tm = TemporalMemory(
cellsPerColumn=4,
connectedPermanence=0.50,
minThreshold=1,
seed=42
)
connections = tm.connections
connections.createSegment(0)
connections.createSynapse(0, 23, 0.6)
connections.createSynapse(0, 37, 0.4)
connections.createSynapse(0, 477, 0.9)
connections.createSegment(0)
connections.createSynapse(1, 49, 0.9)
connections.createSynapse(1, 3, 0.8)
connections.createSegment(1)
connections.createSynapse(2, 733, 0.7)
connections.createSegment(108)
connections.createSynapse(3, 486, 0.9)
activeColumns = set([0, 1, 26])
predictedColumns = set([26])
prevActiveCells = set([23, 37, 49, 733])
prevWinnerCells = set([23, 37, 49, 733])
(activeCells,
winnerCells,
learningSegments) = tm.burstColumns(activeColumns,
predictedColumns,
prevActiveCells,
prevWinnerCells,
connections)
self.assertEqual(activeCells, set([0, 1, 2, 3, 4, 5, 6, 7]))
self.assertEqual(winnerCells, set([0, 6])) # 6 is randomly chosen cell
self.assertEqual(learningSegments, set([0, 4])) # 4 is new segment created
# Check that new segment was added to winner cell (6) in column 1
self.assertEqual(connections.segmentsForCell(6), set([4]))
def testBurstColumnsEmpty(self):
tm = self.tm
activeColumns = set()
predictedColumns = set()
prevActiveCells = set()
prevWinnerCells = set()
connections = tm.connections
(activeCells,
winnerCells,
learningSegments) = tm.burstColumns(activeColumns,
predictedColumns,
prevActiveCells,
prevWinnerCells,
connections)
self.assertEqual(activeCells, set())
self.assertEqual(winnerCells, set())
self.assertEqual(learningSegments, set())
def testLearnOnSegments(self):
tm = TemporalMemory(maxNewSynapseCount=2)
connections = tm.connections
connections.createSegment(0)
connections.createSynapse(0, 23, 0.6)
connections.createSynapse(0, 37, 0.4)
connections.createSynapse(0, 477, 0.9)
connections.createSegment(1)
connections.createSynapse(1, 733, 0.7)
connections.createSegment(8)
connections.createSynapse(2, 486, 0.9)
connections.createSegment(100)
prevActiveSegments = set([0, 2])
learningSegments = set([1, 3])
prevActiveCells = set([23, 37, 733])
winnerCells = set([0])
prevWinnerCells = set([10, 11, 12, 13, 14])
predictedInactiveCells = set()
prevMatchingSegments = set()
tm.learnOnSegments(prevActiveSegments,
learningSegments,
prevActiveCells,
winnerCells,
prevWinnerCells,
connections,
predictedInactiveCells,
prevMatchingSegments)
# Check segment 0
synapseData = connections.dataForSynapse(0)
self.assertAlmostEqual(synapseData.permanence, 0.7)
synapseData = connections.dataForSynapse(1)
self.assertAlmostEqual(synapseData.permanence, 0.5)
synapseData = connections.dataForSynapse(2)
self.assertAlmostEqual(synapseData.permanence, 0.8)
# Check segment 1
synapseData = connections.dataForSynapse(3)
self.assertAlmostEqual(synapseData.permanence, 0.8)
self.assertEqual(len(connections.synapsesForSegment(1)), 2)
# Check segment 2
synapseData = connections.dataForSynapse(4)
self.assertAlmostEqual(synapseData.permanence, 0.9)
self.assertEqual(len(connections.synapsesForSegment(2)), 1)
# Check segment 3
self.assertEqual(len(connections.synapsesForSegment(3)), 2)
def testComputePredictiveCells(self):
tm = TemporalMemory(activationThreshold=2, minThreshold=2)
connections = tm.connections
connections.createSegment(0)
connections.createSynapse(0, 23, 0.6)
connections.createSynapse(0, 37, 0.5)
connections.createSynapse(0, 477, 0.9)
connections.createSegment(1)
connections.createSynapse(1, 733, 0.7)
connections.createSynapse(1, 733, 0.4)
connections.createSegment(1)
connections.createSynapse(2, 974, 0.9)
connections.createSegment(8)
connections.createSynapse(3, 486, 0.9)
connections.createSegment(100)
activeCells = set([23, 37, 733, 974])
(activeSegments,
predictiveCells,
matchingSegments,
matchingCells) = tm.computePredictiveCells(activeCells, connections)
self.assertEqual(activeSegments, set([0]))
self.assertEqual(predictiveCells, set([0]))
self.assertEqual(matchingSegments, set([0,1]))
self.assertEqual(matchingCells, set([0,1]))
def testBestMatchingCell(self):
tm = TemporalMemory(
connectedPermanence=0.50,
minThreshold=1,
seed=42
)
connections = tm.connections
connections.createSegment(0)
connections.createSynapse(0, 23, 0.6)
connections.createSynapse(0, 37, 0.4)
connections.createSynapse(0, 477, 0.9)
connections.createSegment(0)
connections.createSynapse(1, 49, 0.9)
connections.createSynapse(1, 3, 0.8)
connections.createSegment(1)
connections.createSynapse(2, 733, 0.7)
connections.createSegment(108)
connections.createSynapse(3, 486, 0.9)
activeCells = set([23, 37, 49, 733])
self.assertEqual(tm.bestMatchingCell(tm.cellsForColumn(0),
activeCells,
connections),
(0, 0))
self.assertEqual(tm.bestMatchingCell(tm.cellsForColumn(3), # column containing cell 108
activeCells,
connections),
(96, None)) # Random cell from column
self.assertEqual(tm.bestMatchingCell(tm.cellsForColumn(999),
activeCells,
connections),
(31972, None)) # Random cell from column
def testBestMatchingCellFewestSegments(self):
tm = TemporalMemory(
columnDimensions=[2],
cellsPerColumn=2,
connectedPermanence=0.50,
minThreshold=1,
seed=42
)
connections = tm.connections
connections.createSegment(0)
connections.createSynapse(0, 3, 0.3)
activeSynapsesForSegment = set([])
for _ in range(100):
# Never pick cell 0, always pick cell 1
(cell, _) = tm.bestMatchingCell(tm.cellsForColumn(0),
activeSynapsesForSegment,
connections)
self.assertEqual(cell, 1)
def testBestMatchingSegment(self):
tm = TemporalMemory(
connectedPermanence=0.50,
minThreshold=1
)
connections = tm.connections
connections.createSegment(0)
connections.createSynapse(0, 23, 0.6)
connections.createSynapse(0, 37, 0.4)
connections.createSynapse(0, 477, 0.9)
connections.createSegment(0)
connections.createSynapse(1, 49, 0.9)
connections.createSynapse(1, 3, 0.8)
connections.createSegment(1)
connections.createSynapse(2, 733, 0.7)
connections.createSegment(8)
connections.createSynapse(3, 486, 0.9)
activeCells = set([23, 37, 49, 733])
self.assertEqual(tm.bestMatchingSegment(0,
activeCells,
connections),
(0, 2))
self.assertEqual(tm.bestMatchingSegment(1,
activeCells,
connections),
(2, 1))
self.assertEqual(tm.bestMatchingSegment(8,
activeCells,
connections),
(None, None))
self.assertEqual(tm.bestMatchingSegment(100,
activeCells,
connections),
(None, None))
def testLeastUsedCell(self):
tm = TemporalMemory(
columnDimensions=[2],
cellsPerColumn=2,
seed=42
)
connections = tm.connections
connections.createSegment(0)
connections.createSynapse(0, 3, 0.3)
for _ in range(100):
# Never pick cell 0, always pick cell 1
self.assertEqual(tm.leastUsedCell(tm.cellsForColumn(0),
connections),
1)
def testAdaptSegment(self):
tm = self.tm
connections = tm.connections
connections.createSegment(0)
connections.createSynapse(0, 23, 0.6)
connections.createSynapse(0, 37, 0.4)
connections.createSynapse(0, 477, 0.9)
tm.adaptSegment(0, set([0, 1]), connections,
tm.permanenceIncrement,
tm.permanenceDecrement)
synapseData = connections.dataForSynapse(0)
self.assertAlmostEqual(synapseData.permanence, 0.7)
synapseData = connections.dataForSynapse(1)
self.assertAlmostEqual(synapseData.permanence, 0.5)
synapseData = connections.dataForSynapse(2)
self.assertAlmostEqual(synapseData.permanence, 0.8)
def testAdaptSegmentToMax(self):
tm = self.tm
connections = tm.connections
connections.createSegment(0)
connections.createSynapse(0, 23, 0.9)
tm.adaptSegment(0, set([0]), connections,
tm.permanenceIncrement,
tm.permanenceDecrement)
synapseData = connections.dataForSynapse(0)
self.assertAlmostEqual(synapseData.permanence, 1.0)
# Now permanence should be at max
tm.adaptSegment(0, set([0]), connections,
tm.permanenceIncrement,
tm.permanenceDecrement)
synapseData = connections.dataForSynapse(0)
self.assertAlmostEqual(synapseData.permanence, 1.0)
def testAdaptSegmentToMin(self):
tm = self.tm
connections = tm.connections
connections.createSegment(0)
connections.createSynapse(0, 23, 0.1)
tm.adaptSegment(0, set(), connections,
tm.permanenceIncrement,
tm.permanenceDecrement)
synapseData = connections.dataForSynapse(0)
self.assertAlmostEqual(synapseData.permanence, 0.0)
# Now permanence should be at min
tm.adaptSegment(0, set(), connections,
tm.permanenceIncrement,
tm.permanenceDecrement)
synapseData = connections.dataForSynapse(0)
self.assertAlmostEqual(synapseData.permanence, 0.0)
def testPickCellsToLearnOn(self):
tm = TemporalMemory(seed=42)
connections = tm.connections
connections.createSegment(0)
winnerCells = set([4, 47, 58, 93])
self.assertEqual(tm.pickCellsToLearnOn(2, 0, winnerCells, connections),
set([4, 58])) # randomly picked
self.assertEqual(tm.pickCellsToLearnOn(100, 0, winnerCells, connections),
set([4, 47, 58, 93]))
self.assertEqual(tm.pickCellsToLearnOn(0, 0, winnerCells, connections),
set())
def testPickCellsToLearnOnAvoidDuplicates(self):
tm = TemporalMemory(seed=42)
connections = tm.connections
connections.createSegment(0)
connections.createSynapse(0, 23, 0.6)
winnerCells = set([23])
# Ensure that no additional (duplicate) cells were picked
self.assertEqual(tm.pickCellsToLearnOn(2, 0, winnerCells, connections),
set())
def testColumnForCell1D(self):
tm = TemporalMemory(
columnDimensions=[2048],
cellsPerColumn=5
)
self.assertEqual(tm.columnForCell(0), 0)
self.assertEqual(tm.columnForCell(4), 0)
self.assertEqual(tm.columnForCell(5), 1)
self.assertEqual(tm.columnForCell(10239), 2047)
def testColumnForCell2D(self):
tm = TemporalMemory(
columnDimensions=[64, 64],
cellsPerColumn=4
)
self.assertEqual(tm.columnForCell(0), 0)
self.assertEqual(tm.columnForCell(3), 0)
self.assertEqual(tm.columnForCell(4), 1)
self.assertEqual(tm.columnForCell(16383), 4095)
def testColumnForCellInvalidCell(self):
tm = TemporalMemory(
columnDimensions=[64, 64],
cellsPerColumn=4
)
try:
tm.columnForCell(16383)
except IndexError:
self.fail("IndexError raised unexpectedly")
args = [16384]
self.assertRaises(IndexError, tm.columnForCell, *args)
args = [-1]
self.assertRaises(IndexError, tm.columnForCell, *args)
def testCellsForColumn1D(self):
tm = TemporalMemory(
columnDimensions=[2048],
cellsPerColumn=5
)
expectedCells = set([5, 6, 7, 8, 9])
self.assertEqual(tm.cellsForColumn(1), expectedCells)
def testCellsForColumn2D(self):
tm = TemporalMemory(
columnDimensions=[64, 64],
cellsPerColumn=4
)
expectedCells = set([256, 257, 258, 259])
self.assertEqual(tm.cellsForColumn(64), expectedCells)
def testCellsForColumnInvalidColumn(self):
tm = TemporalMemory(
columnDimensions=[64, 64],
cellsPerColumn=4
)
try:
tm.cellsForColumn(4095)
except IndexError:
self.fail("IndexError raised unexpectedly")
args = [4096]
self.assertRaises(IndexError, tm.cellsForColumn, *args)
args = [-1]
self.assertRaises(IndexError, tm.cellsForColumn, *args)
def testNumberOfColumns(self):
tm = TemporalMemory(
columnDimensions=[64, 64],
cellsPerColumn=32
)
self.assertEqual(tm.numberOfColumns(), 64 * 64)
def testNumberOfCells(self):
tm = TemporalMemory(
columnDimensions=[64, 64],
cellsPerColumn=32
)
self.assertEqual(tm.numberOfCells(), 64 * 64 * 32)
def testMapCellsToColumns(self):
tm = TemporalMemory(
columnDimensions=[100],
cellsPerColumn=4
)
columnsForCells = tm.mapCellsToColumns(set([0, 1, 2, 5, 399]))
self.assertEqual(columnsForCells[0], set([0, 1, 2]))
self.assertEqual(columnsForCells[1], set([5]))
self.assertEqual(columnsForCells[99], set([399]))
def testWrite(self):
tm1 = TemporalMemory(
columnDimensions=[100],
cellsPerColumn=4,
activationThreshold=7,
initialPermanence=0.37,
connectedPermanence=0.58,
minThreshold=4,
maxNewSynapseCount=18,
permanenceIncrement=0.23,
permanenceDecrement=0.08,
seed=91
)
# Run some data through before serializing
self.patternMachine = PatternMachine(100, 4)
self.sequenceMachine = SequenceMachine(self.patternMachine)
sequence = self.sequenceMachine.generateFromNumbers(range(5))
for _ in range(3):
for pattern in sequence:
tm1.compute(pattern)
proto1 = TemporalMemoryProto_capnp.TemporalMemoryProto.new_message()
tm1.write(proto1)
# Write the proto to a temp file and read it back into a new proto
with tempfile.TemporaryFile() as f:
proto1.write(f)
f.seek(0)
proto2 = TemporalMemoryProto_capnp.TemporalMemoryProto.read(f)
# Load the deserialized proto
tm2 = TemporalMemory.read(proto2)
# Check that the two temporal memory objects have the same attributes
self.assertEqual(tm1, tm2)
# Run a couple records through after deserializing and check results match
tm1.compute(self.patternMachine.get(0))
tm2.compute(self.patternMachine.get(0))
self.assertEqual(tm1.activeCells, tm2.activeCells)
self.assertEqual(tm1.predictiveCells, tm2.predictiveCells)
self.assertEqual(tm1.winnerCells, tm2.winnerCells)
self.assertEqual(tm1.connections, tm2.connections)
tm1.compute(self.patternMachine.get(3))
tm2.compute(self.patternMachine.get(3))
self.assertEqual(tm1.activeCells, tm2.activeCells)
self.assertEqual(tm1.predictiveCells, tm2.predictiveCells)
self.assertEqual(tm1.winnerCells, tm2.winnerCells)
self.assertEqual(tm1.connections, tm2.connections)
