def forward(network, x):
W1, W2, W3 = network['W1'], network['W2'], network['W3']
b1, b2, b3 = network['b1'], network['b2'], network['b3']
a1 = np.dot(x, W1) + b1
z1 = sigmoid(a1)
a2 = np.dot(z1, W2) + b2
z2 = sigmoid(a2)
a3 = np.dot(z2, W3) + b3
y = identity(a3)
return y
def predict(network, x):
W1, W2, W3 = network['W1'], network['W2'], network['W3']
b1, b2, b3 = network['b1'], network['b2'], network['b3']
a1 = np.dot(x, W1) + b1
z1 = sigmoid(a1)
a2 = np.dot(z1, W2) + b2
z2 = sigmoid(a2)
a3 = np.dot(z2, W3) + b3
y = softmax(a3)
return y
def get_probability_visible(self, visibles, hidden):
"""
Returns the probability of setting visible units to 1, given the
the hidden units and history of visible units.
Assume visibles[0] = V{t-1}, visibles[1] = V{t-2}, etc.
Then hiddens = H{t}
"""
B_V = self.get_bias_visible(visibles)
return sigmoid(np.dot(self.weights, hidden) + B_V)
def get_probability_hidden(self, visibles, hiddens):
"""
Returns the probability of setting hidden units to 1, given the
history of hidden and visible units.
Assumes visibles[0] = V{t}, visibles[1] = V{t-1}, ...
and hiddens[0] = H{t-1}, hiddens[1] = H{t-2}, ...
visibles should have one extra entry than hiddens
"""
# H = sigmoid(W'V_t + B_H(V_{t-m...t-1}, H{t-m...t-1}))
B_H = self.get_bias_hidden(visibles[1:], hiddens)
return sigmoid(np.dot(self.weights.transpose(), visibles[0]) + B_H)
