def test_out_prediction():
setting = SpatialSettings(debug = False, min_overlap = 1, desired_local_activity = 4, connected_pct = 1,
connected_perm = 0.01, xinput = 3, yinput = 3, potential_radius = 4, xdimension = 3,
ydimension = 3, initial_inhibition_radius = 1, permanence_inc = 0.1, permanence_dec = 0.1,
max_boost = 2, min_duty_cycle_fraction = 0.2)
setting.debug = True
setting.min_overlap = 1
setting.desired_local_activity = 1
setting.connected_pct = 1
setting.xinput = 4
setting.yinput = 4
setting.potential_radius = 2
setting.xdimension = 4
setting.ydimension = 1
setting.initial_inhibition_radius = 2
r = Region(setting, SimpleMapper())
inp = [[1, 0, 1, 0], [1, 0, 1, 0], [1, 0, 1, 0], [1, 0, 1, 0]]
r.step_forward(inp)
res = r.out_prediction([[1, 0, 1, 0]])
# print(res)
for i in range(len(res)):
for j in range(len(res[0])):
if inp[i][j] ==1:
assert res[i][j] > 0
def test_htm_constructuion():
setting = SpatialSettings(debug = False, min_overlap = 1, desired_local_activity = 4, connected_pct = 1,
connected_perm = 0.01, xinput = 3, yinput = 3, potential_radius = 4, xdimension = 3,
ydimension = 3, initial_inhibition_radius = 1, permanence_inc = 0.1, permanence_dec = 0.1,
max_boost = 2, min_duty_cycle_fraction = 0.2)
setting.debug = True
setting.min_overlap = 1
setting.desired_local_activity = 1
setting.connected_pct = 1
setting.xinput = 5
setting.yinput = 1
setting.potential_radius = 2
setting.xdimension = 4
setting.ydimension = 1
setting.initial_inhibition_radius = 2
r = Region(setting, SimpleMapper())
assert len(r.get_columns()) == 4
assert r.get_input_h() == 1
assert r.get_input_w() == 5
assert len(r.get_columns()[0].get_neighbors()) == 2
v = r.get_columns()[r.get_columns()[0].get_neighbors()[0]].get_coord()
assert v[0] == 1.0 and v[1] == 0.0
def test_overlap_on_ones():
inp = [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1]
settings = SpatialSettings(debug = False, min_overlap = 1, desired_local_activity = 4, connected_pct = 1,
connected_perm = 0.01, xinput = 3, yinput = 3, potential_radius = 4, xdimension = 3,
ydimension = 3, initial_inhibition_radius = 1, permanence_inc = 0.1, permanence_dec = 0.1,
max_boost = 2, min_duty_cycle_fraction = 0.2)
settings.debug = True
settings.min_overlap = 1
settings.desired_local_activity = 1
settings.connected_pct = 1
settings.xinput = len(inp)
settings.yinput = 1
settings.potential_radius = 2
settings.xdimension = 4
settings.ydimension = 1
r = Region(settings, SimpleMapper())
overlaps = r.update_overlaps(r.get_columns(), inp)
groundtruth = [5, 5, 5, 5]
for i in range(len(groundtruth)):
assert overlaps[i] == groundtruth[i]
settings.potential_radius = 2
settings.xdimension = 1
settings.ydimension = 1
r = Region(settings, SimpleMapper())
overlaps = r.update_overlaps(r.get_columns(), inp)
groundtruth = [5]
for i in range(len(groundtruth)):
assert overlaps[i] == groundtruth[i]
settings.potential_radius = 2
settings.xdimension = 16
settings.ydimension = 1
r = Region(settings, SimpleMapper())
overlaps = r.update_overlaps(r.get_columns(), inp)
groundtruth = [3, 4, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 3]
for i in range(len(groundtruth)):
assert overlaps[i] == groundtruth[i]
def test_debug_false():
setting = SpatialSettings(debug = False, min_overlap = 1, desired_local_activity = 4, connected_pct = 1,
connected_perm = 0.01, xinput = 3, yinput = 3, potential_radius = 4, xdimension = 3,
ydimension = 3, initial_inhibition_radius = 1, permanence_inc = 0.1, permanence_dec = 0.1,
max_boost = 2, min_duty_cycle_fraction = 0.2)
setting.debug = False
setting.min_overlap = 1
setting.desired_local_activity = 1
setting.connected_pct = 1
setting.xinput = 4
setting.yinput = 4
setting.potential_radius = 2
setting.xdimension = 4
setting.ydimension = 1
setting.initial_inhibition_radius = 2
r = Region(setting, SimpleMapper())
inp = [[1, 0, 1, 0], [1, 0, 1, 0], [1, 0, 1, 0], [1, 0, 1, 0]]
res = r.step_forward(inp)
print(res)
assert len(list(filter(lambda x: x[0] == True, res)))>0
def test_ladder():
# FileOutputStream fos=new FileOutputStream("out.txt")
# PrintWriter pw=new PrintWriter(fos)
# FileOutputStream fos_in=new FileOutputStream("in.txt")
# PrintWriter pw_in=new PrintWriter(fos_in)
w = 15
h = 15
begx = 0
begy = 0
step_size = 5
map = [[0 for j in range(h)] for i in range(w)]
inp = [0 for i in range(h * w)]
STEPS = 5
TOTAL_STEPS = 1000
STEP_SIZE = STEPS
setting = SpatialSettings(debug = False, min_overlap = 1, desired_local_activity = 4, connected_pct = 1,
connected_perm = 0.01, xinput = 3, yinput = 3, potential_radius = 4, xdimension = 3,
ydimension = 3, initial_inhibition_radius = 1, permanence_inc = 0.1, permanence_dec = 0.1,
max_boost = 2, min_duty_cycle_fraction = 0.2)
setting.debug = True
setting.min_overlap = 1
setting.desired_local_activity = 1
setting.connected_pct = 1
# setting.connectedPerm=0.01
setting.xinput = w
setting.yinput = h
setting.potential_radius = 2
setting.xdimension = 3
setting.ydimension = 3
setting.initial_inhibition_radius = 2
# pw.print(setting.xDimension + " ")
# pw.print(setting.yDimension + " ")
# pw.print(TOTAL_STEPS)
# pw.println()
#
# pw_in.print(setting.xDimension + " ")
# pw_in.print(setting.yDimension + " ")
# pw_in.print(TOTAL_STEPS)
# pw_in.println()
r = Region(setting, VerySimpleMapper())
x = begx
y = begy
for i in range(x, x + step_size):
for j in range(y, y + step_size):
map[i][j] = 1
for step in range(TOTAL_STEPS):
print("DATA:\n")
index = 0
for k in range(w):
for m in range(h):
inp[index] = map[k][m]
# pw_in.print(in[index])
print(str(inp[index]) + " ", end="", flush=True)
index += 1
print()
# pw_in.println()
print()
# pw_in.println()
for i in range(x, x + step_size):
for j in range(y, y + step_size):
if i < len(map) and j < len(map[0]):
map[i][j] = 0
x = x + STEP_SIZE
y = y + STEP_SIZE
if x > w:
x = 0
y = 0
for i in range(x, x + step_size):
for j in range(y, y + step_size):
if i < len(map) and j < len(map[0]):
map[i][j] = 1
for c in r.get_columns():
c.set_is_active(False)
ov = r.update_overlaps(r.get_columns(), inp)
r.inhibition_phase(r.get_columns(), ov)
r.learning_phase(r.get_columns(), inp, ov)
cols = r.get_columns()
for i in range(setting.xdimension):
for j in range(setting.ydimension):
state = 1 if find_by_colxy(cols, i, j).get_is_active() else 0
print(str(state) + " ", end="", flush=True)
# pw.print(state)
# pw.print(" ")
print()
def test_inhibition_phase():
inp = [1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0]
settings = SpatialSettings(debug = False, min_overlap = 1, desired_local_activity = 4, connected_pct = 1,
connected_perm = 0.01, xinput = 3, yinput = 3, potential_radius = 4, xdimension = 3,
ydimension = 3, initial_inhibition_radius = 1, permanence_inc = 0.1, permanence_dec = 0.1,
max_boost = 2, min_duty_cycle_fraction = 0.2)
settings.debug = True
settings.min_overlap = 1
settings.desired_local_activity = 1
settings.connected_pct = 1
settings.xinput = len(inp)
settings.yinput = 1
settings.potential_radius = 2
settings.xdimension = 4
settings.ydimension = 1
settings.initial_inhibition_radius = 1
r = Region(settings, SimpleMapper())
overlaps = r.update_overlaps(r.get_columns(), inp)
cols = r.inhibition_phase(r.get_columns(), overlaps)
assert len(cols) == 2
r = Region(settings, SimpleMapper())
overlaps = r.update_overlaps(r.get_columns(), inp)
r.inhibition_phase(r.get_columns(), overlaps)
# ожидаем разные результаты теста из-за рандомного шафла
cols = r.inhibition_phase(r.get_columns(), overlaps)
assert len(cols) == 2
cols = r.inhibition_phase(r.get_columns(), overlaps)
assert len(cols) == 2
cols = r.inhibition_phase(r.get_columns(), overlaps)
assert len(cols) == 2
cols = r.inhibition_phase(r.get_columns(), overlaps)
assert len(cols) == 2
def testUpdateSynapses():
inp = [1, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1]
settings = SpatialSettings(debug = False, min_overlap = 1, desired_local_activity = 4, connected_pct = 1,
connected_perm = 0.01, xinput = 3, yinput = 3, potential_radius = 4, xdimension = 3,
ydimension = 3, initial_inhibition_radius = 1, permanence_inc = 0.1, permanence_dec = 0.1,
max_boost = 2, min_duty_cycle_fraction = 0.2)
settings.debug = True
settings.min_overlap = 1
settings.desired_local_activity = 1
settings.connected_pct = 1
settings.xinput = len(inp)
settings.yinput = 1
settings.potential_radius = 2
settings.xdimension = 4
settings.ydimension = 1
settings.initial_inhibition_radius = 1
settings.permanence_inc = 0.2
settings.permanence_dec = 0.2
r = Region(settings, SimpleMapper())
overlaps = r.update_overlaps(r.get_columns(), inp)
r.inhibition_phase(r.get_columns(), overlaps)
r.get_columns()[0].get_potential_synapses().get(4).set_permanence(0.5)
r.update_synapses(r.get_columns(), inp)
v = r.get_columns()[0].get_potential_synapses().get(4).get_permanence()
assert v == 0.7
r.get_columns()[0].get_potential_synapses().get(5).set_permanence(0.5)
r.update_synapses(r.get_columns(), inp)
v = r.get_columns()[0].get_potential_synapses().get(5).get_permanence()
assert v == 0.3
def test_update_active_duty_cycle():
inp = [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1]
settings = SpatialSettings(debug = False, min_overlap = 1, desired_local_activity = 4, connected_pct = 1,
connected_perm = 0.01, xinput = 3, yinput = 3, potential_radius = 4, xdimension = 3,
ydimension = 3, initial_inhibition_radius = 1, permanence_inc = 0.1, permanence_dec = 0.1,
max_boost = 2, min_duty_cycle_fraction = 0.2)
settings.debug = True
settings.min_overlap = 1
settings.desired_local_activity = 1
settings.connected_pct = 1
settings.xinput = len(inp)
settings.yinput = 1
settings.potential_radius = 2
settings.xdimension = 4
settings.ydimension = 1
settings.initial_inhibition_radius = 1
r = Region(settings, SimpleMapper())
overlaps = r.update_overlaps(r.get_columns(), inp)
r.inhibition_phase(r.get_columns(), overlaps)
r.update_active_duty_cycle(r.get_columns())
r.inhibition_phase(r.get_columns(), overlaps)
r.update_active_duty_cycle(r.get_columns())
r.inhibition_phase(r.get_columns(), overlaps)
r.update_active_duty_cycle(r.get_columns())
r.inhibition_phase(r.get_columns(), overlaps)
r.update_active_duty_cycle(r.get_columns())
assert len(r.get_active_duty_cycles()) == len(r.get_columns())
assert r.get_active_duty_cycles()[0] == 4
inp = [1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0]
r = Region(settings, SimpleMapper())
overlaps = r.update_overlaps(r.get_columns(), inp)
r.inhibition_phase(r.get_columns(), overlaps)
r.update_active_duty_cycle(r.get_columns())
r.inhibition_phase(r.get_columns(), overlaps)
r.update_active_duty_cycle(r.get_columns())
r.inhibition_phase(r.get_columns(), overlaps)
r.update_active_duty_cycle(r.get_columns())
r.inhibition_phase(r.get_columns(), overlaps)
r.update_active_duty_cycle(r.get_columns())
assert len(r.get_active_duty_cycles()) == len(r.get_columns())
assert r.get_active_duty_cycles()[0] == 0
assert r.get_active_duty_cycles()[1] == 3
assert r.get_active_duty_cycles()[2] == 3
assert r.get_active_duty_cycles()[3] == 2
from spatialPooler.sp_region import Region as SpatialPoolerRegion
from apps.settings import *
input_settings = InputSettings(SCALE=1, STEPS_NUMBER=300)
generator = ListImages(12)
spatial_settings = SpatialSettings(debug = False, min_overlap = 1, desired_local_activity = 2, connected_pct = 1,
connected_perm = 0.1, xinput = generator.square_size, yinput = generator.square_size,
potential_radius = 4, xdimension = 12, ydimension = 12, initial_inhibition_radius = 2,
permanence_inc = 0.02, permanence_dec = 0.1, max_boost = 2, min_duty_cycle_fraction = 0.2)
temporal_settings = TemporalSettings(region_size=spatial_settings.xdimension, column_size=4, initial_permanence=0.5,
dendrite_activate_threshold=1, dendrite_permanence_inc_delta=0.02,
dendrite_permanence_dec_delta=-0.1, passive_time_to_active_threshold=1000,
synapse_threshold=0.45)
spatial_pooler_region = SpatialPoolerRegion(spatial_settings=spatial_settings, mapper=SquareMapperAutoRadius())
temporal_pooler_region = TemporalPoolerRegion(temporal_settings=temporal_settings)
for i in range(input_settings.STEPS_NUMBER):
generator.out()
outdata=spatial_pooler_region.step_forward(generator.get_data())
print_matrix(to_binmatrix(outdata))
temporal_pooler_region.step_forward(outdata)
temporal_pooler_region.out_prediction()
cols=temporal_pooler_region.columns
imagined_input = spatial_pooler_region.out_prediction(cols, temporal_settings.region_size)
generator.move()
