def testCreatingLinks(self):
l = HTMLevel(0, {"size": [1], "nodeCloning": True}, previous_level_size=[1])
prev_size = [8, 8]
correct_links = {0: range(np.prod(prev_size))}
links = l._get_links([1, 1], prev_size)
self.assertEqual(links, correct_links)
def testCreatingLinks(self):
l = HTMLevel(0, {
'size': [1],
'nodeCloning': True
},
previous_level_size=[1])
prev_size = [8, 8]
correct_links = {0: range(np.prod(prev_size))}
links = l._get_links([1, 1], prev_size)
self.assertEqual(links, correct_links)
def runTest(self):
params = {"transitionMemory": 4, "requestedGroupsCount": 3}
l = HTMLevel(0, {"size": [1], "nodeCloning": True}, previous_level_size=[1])
for p in patterns:
l.do_spatial_learning([p])
for p in patterns:
l.do_temporal_learning([p])
desired_TAM = np.array([[2.0, 4.0, 3.0, 5.0], [3.0, 0.0, 4.0, 2.0], [4.0, 0.0, 0.0, 3.0], [0.0, 0.0, 0.0, 0.0]])
self.assertEqual(np.array_equal(l.nodes[0].tp.TAM, desired_TAM), True)
l.finalize_learning()
self.assertEqual(len(l.nodes[0].tp.temporal_groups), params["requestedGroupsCount"])
# testing creating nodes using params
params = {
"nodeCloning": True,
"size": [8, 8],
"overlap": [0, 0],
# Spatial pooler
"maxCoincidenceCount": 128,
"spatialPoolerAlgorithm": "gaussian",
"sigma": 1,
"maxDistance": 0.1,
# Temporal pooler
"requestedGroupsCount": 20,
"temporalPoolerAlgorithm": "maxProp",
"transitionMemory": 4,
}
l = HTMLevel(1, params, previous_level_size=[1])
self.assertEqual(len(l.nodes), 1)
def runTest(self):
params = {'transitionMemory': 4, 'requestedGroupsCount': 3}
l = HTMLevel(0, {
'size': [1],
'nodeCloning': True
},
previous_level_size=[1])
for p in patterns:
l.do_spatial_learning([p])
for p in patterns:
l.do_temporal_learning([p])
desired_TAM = np.array([[2., 4., 3., 5.], [3., 0., 4., 2.],
[4., 0., 0., 3.], [0., 0., 0., 0.]])
self.assertEqual(np.array_equal(l.nodes[0].tp.TAM, desired_TAM), True)
l.finalize_learning()
self.assertEqual(len(l.nodes[0].tp.temporal_groups),
params['requestedGroupsCount'])
# testing creating nodes using params
params = {
'nodeCloning': True,
'size': [8, 8],
'overlap': [0, 0],
# Spatial pooler
'maxCoincidenceCount': 128,
'spatialPoolerAlgorithm': 'gaussian',
'sigma': 1,
'maxDistance': 0.1,
# Temporal pooler
'requestedGroupsCount': 20,
'temporalPoolerAlgorithm': 'maxProp',
'transitionMemory': 4,
}
l = HTMLevel(1, params, previous_level_size=[1])
self.assertEqual(len(l.nodes), 1)
def create_network(self, params):
"""Create network based on the provided parameters.
Args:
params: list of dicts sorted starting from the bottom level
"""
if params is None:
raise ValueError("Parameter 'params' should not be empty.")
if self.verbose: print("Creating network...")
for level_no in range(len(params)):
if level_no == 0:
# assuming rectangular arrangement of nodes
if self.is_image_sensor:
self.patch_size = [
int(self.sensor.width / params[0]['size'][0]),
int(self.sensor.height / params[0]['size'][1])
]
if self.verbose:
print("Patch size is set to %s." % self.patch_size)
prev_level_size = params[0]['size']
else:
prev_level_size = params[level_no - 1]['size']
level = HTMLevel(level_no,
params=params[level_no],
debug=self.debug,
previous_level_size=prev_level_size)
self.add_level(level)
for level_no in range(len(self.levels) - 1):
self.levels[level_no].above_links = self.levels[level_no + 1].links
if self.verbose: print("=" * 40)
