def test_train(self):
"""
Test forward computation
"""
layer = SpatialPooler(4,
learning_rate=0.1,
pool_density=1,
boost_strength=0)
x = tf.placeholder(tf.float32, [None, 4], name='Input')
layer.build([1, 4])
train = layer.train(x, tf.constant([[0., 1, 0, 1]]))
with tf.Session() as sess:
sess.run(tf.global_variables_initializer())
# Override permenance with a custom value
sess.run(
tf.assign(layer.permanence,
[[0.6, 0, 0.6, 0.6], [0, 0.6, 0, 0],
[0.6, 0.6, 0.6, 0], [0, 0, 0.6, 0.6]]))
# Compute
result = sess.run(train, {x: [[1, 1, 0, 1]]})
# Check the new permenance
test.assert_allclose(result[0],
[[0.6, 0, 0.6, 0.7], [0, 0.7, 0, 0],
[0.6, 0.5, 0.6, 0], [0, 0, 0.6, 0.7]])
def __init__(self):
pooler = SpatialPooler(htm_units, lr=1e-2)
# Model input
self.x = tf.placeholder(tf.float32, [None, input_units])
self.y = pooler(self.x)
self.train_ops = pooler.train_ops
cluster = RecruitmentLayer(htm_units, num_clusters=6)
self.x1 = tf.placeholder(tf.float32, [1, htm_units])
self.y1 = cluster(self.x1)
self.recruitment_train_ops = cluster.train_ops
def __init__(self):
pooler = SpatialPooler(htm_units, lr=1e-2)
# Model input
self.x = tf.placeholder(tf.float32, [None, input_units])
self.y = pooler(self.x)
self.train_ops = pooler.train_ops
# Build classifier
classifier_in = Input((htm_units, ))
classifier_out = Dense(num_classes,
activation='softmax')(classifier_in)
self.classifier = Model(classifier_in, classifier_out)
self.classifier.compile(optimizer='adam',
loss='categorical_crossentropy',
metrics=['acc'])
def __init__(self, input_units, sp_units, clusters):
self.input_units = input_units
self.sp_units = sp_units
self.clusters = clusters
self.pooler = SpatialPooler(sp_units, lr=1e-2)
# Model input
self.x = tf.placeholder(tf.float32, [None, input_units])
self.y = self.pooler(self.x)
self.train_ops = self.pooler.train_ops
cluster = RecruitmentLayer(sp_units, num_clusters=clusters)
self.x1 = tf.placeholder(tf.float32, [1, sp_units])
self.y1 = cluster(self.x1)
self.recruitment_train_ops = cluster.train_ops
def test_call(self):
"""
Test forward computation
"""
layer = SpatialPooler(4, pool_density=1, boost_strength=0)
x = tf.placeholder(tf.float32, [None, 4], name='Input')
y = layer(x)
with tf.Session() as sess:
sess.run(tf.global_variables_initializer())
# Override permenance with a custom value
sess.run(
tf.assign(
layer.permanence,
[[1, 0, 1, 1], [0, 1, 0, 0], [1, 1, 1, 0], [0, 0, 1, 1]]))
# Compute
result = sess.run(y, {x: [[1, 1, 0, 1], [1, 1, 0, 0]]})
test.assert_array_equal(result[0], [0, 0, 1, 0])
def __init__(self):
pooler = SpatialPooler(htm_units, lr=1e-2)
# Model input
self.x = tf.placeholder(tf.float32, [None, input_units])
self.y = pooler(self.x)
self.train_ops = pooler.train_ops