class AbstractSensorimotorTest(unittest.TestCase):
VERBOSITY = 1
DEFAULT_TM_PARAMS = {}
SEED = 42
def _init(self, tmOverrides=None):
"""
Initialize Sensorimotor Temporal Memory, and other member variables.
:param tmOverrides: overrides for default Temporal Memory parameters
"""
params = self._computeTMParams(tmOverrides)
# Uncomment the line below to disable learn on one cell mode
# params["learnOnOneCell"] = False
self.tm = SensorimotorTemporalMemory(**params)
def _feedTM(self, sequence, learn=True):
(sensorSequence, motorSequence, sensorimotorSequence,
sequenceLabels) = sequence
self.tm.clearHistory()
for i in xrange(len(sensorSequence)):
sensorPattern = sensorSequence[i]
sensorimotorPattern = sensorimotorSequence[i]
sequenceLabel = sequenceLabels[i]
if sensorPattern is None:
self.tm.reset()
else:
self.tm.compute(sensorPattern,
activeExternalCells=sensorimotorPattern,
formInternalConnections=False,
learn=learn,
sequenceLabel=sequenceLabel)
if self.VERBOSITY >= 2:
print self.tm.prettyPrintHistory(verbosity=self.VERBOSITY - 2)
print
if self.VERBOSITY >= 2:
print self.tm.prettyPrintSequenceCellRepresentations()
print
if learn and self.VERBOSITY >= 3:
print self.tm.prettyPrintConnections()
def _testTM(self, sequence):
self._feedTM(sequence, learn=False)
stats = self.tm.getStatistics()
self.allStats.append((self.id(), stats))
return stats
# ==============================
# Overrides
# ==============================
@classmethod
def setUpClass(cls):
cls.allStats = []
@classmethod
def tearDownClass(cls):
cols = [
"Test", "# of predicted active cells (stats)",
"# of predicted inactive cells (stats)",
"# of predicted active columns (stats)",
"# of predicted inactive columns (stats)",
"# of unpredicted active columns (stats)",
"# of predicted active cells per column in each sequence (stats)",
"# of sequences each predicted active cell shows up in"
]
table = PrettyTable(cols)
for stats in cls.allStats:
row = [stats[0]] + [tuple(x) for x in list(stats[1])]
table.add_row(row)
print
print table
print "(stats) => (min, max, sum, average, standard deviation)"
def setUp(self):
self.tm = None
self._random = numpy.random.RandomState(self.SEED)
if self.VERBOSITY >= 2:
print(
"\n"
"======================================================\n"
"Test: {0} \n"
"{1}\n"
"======================================================\n"
).format(self.id(), self.shortDescription())
# ==============================
# Helper functions
# ==============================
@classmethod
def _generateSensorimotorSequences(cls, length, agents):
"""
@param length (int) Length of each sequence to generate, one for
each agent
@param agents (AbstractAgent) Agents acting in their worlds
@return (tuple) (sensor sequence, motor sequence, sensorimotor sequence,
sequence labels)
"""
sensorSequence = []
motorSequence = []
sensorimotorSequence = []
sequenceLabels = []
for agent in agents:
s, m, sm = agent.generateSensorimotorSequence(
length, verbosity=cls.VERBOSITY - 1)
sensorSequence += s
motorSequence += m
sensorimotorSequence += sm
sequenceLabels += [str(agent.world)] * length
sensorSequence.append(None)
motorSequence.append(None)
sensorimotorSequence.append(None)
sequenceLabels.append(None)
return (sensorSequence, motorSequence, sensorimotorSequence,
sequenceLabels)
def _computeTMParams(self, overrides):
params = dict(self.DEFAULT_TM_PARAMS)
params.update(overrides or {})
return params
class AbstractSensorimotorTest(unittest.TestCase):
VERBOSITY = 1
DEFAULT_TM_PARAMS = {}
SEED = 42
def _init(self, tmOverrides=None):
"""
Initialize Sensorimotor Temporal Memory, and other member variables.
:param tmOverrides: overrides for default Temporal Memory parameters
"""
params = self._computeTMParams(tmOverrides)
# Uncomment the line below to disable learn on one cell mode
# params["learnOnOneCell"] = False
self.tm = SensorimotorTemporalMemory(**params)
def _feedTM(self, sequence, learn=True):
(sensorSequence,
motorSequence,
sensorimotorSequence,
sequenceLabels) = sequence
self.tm.clearHistory()
for i in xrange(len(sensorSequence)):
sensorPattern = sensorSequence[i]
sensorimotorPattern = sensorimotorSequence[i]
sequenceLabel = sequenceLabels[i]
if sensorPattern is None:
self.tm.reset()
else:
self.tm.compute(sensorPattern,
activeExternalCells=sensorimotorPattern,
formInternalConnections=False,
learn=learn,
sequenceLabel=sequenceLabel)
if self.VERBOSITY >= 2:
print self.tm.prettyPrintHistory(verbosity=self.VERBOSITY-2)
print
if self.VERBOSITY >= 2:
print self.tm.prettyPrintSequenceCellRepresentations()
print
if learn and self.VERBOSITY >= 3:
print self.tm.prettyPrintConnections()
def _testTM(self, sequence):
self._feedTM(sequence, learn=False)
stats = self.tm.getStatistics()
self.allStats.append((self.id(), stats))
return stats
# ==============================
# Overrides
# ==============================
@classmethod
def setUpClass(cls):
cls.allStats = []
@classmethod
def tearDownClass(cls):
cols = ["Test",
"# of predicted active cells (stats)",
"# of predicted inactive cells (stats)",
"# of predicted active columns (stats)",
"# of predicted inactive columns (stats)",
"# of unpredicted active columns (stats)",
"# of predicted active cells per column in each sequence (stats)",
"# of sequences each predicted active cell shows up in"]
table = PrettyTable(cols)
for stats in cls.allStats:
row = [stats[0]] + [tuple(x) for x in list(stats[1])]
table.add_row(row)
print
print table
print "(stats) => (min, max, sum, average, standard deviation)"
def setUp(self):
self.tm = None
self._random = numpy.random.RandomState(self.SEED)
if self.VERBOSITY >= 2:
print ("\n"
"======================================================\n"
"Test: {0} \n"
"{1}\n"
"======================================================\n"
).format(self.id(), self.shortDescription())
# ==============================
# Helper functions
# ==============================
@classmethod
def _generateSensorimotorSequences(cls, length, agents):
"""
@param length (int) Length of each sequence to generate, one for
each agent
@param agents (AbstractAgent) Agents acting in their worlds
@return (tuple) (sensor sequence, motor sequence, sensorimotor sequence,
sequence labels)
"""
sensorSequence = []
motorSequence = []
sensorimotorSequence = []
sequenceLabels = []
for agent in agents:
s,m,sm = agent.generateSensorimotorSequence(length,
verbosity=cls.VERBOSITY-1)
sensorSequence += s
motorSequence += m
sensorimotorSequence += sm
sequenceLabels += [str(agent.world)] * length
sensorSequence.append(None)
motorSequence.append(None)
sensorimotorSequence.append(None)
sequenceLabels.append(None)
return (sensorSequence, motorSequence, sensorimotorSequence, sequenceLabels)
def _computeTMParams(self, overrides):
params = dict(self.DEFAULT_TM_PARAMS)
params.update(overrides or {})
return params
