def __forward_propagation(self, X):
W1 = self.paramters['W1']
b1 = self.paramters['b1']
W2 = self.paramters['W2']
b2 = self.paramters['b2']
# forward propagation
Z1 = np.dot(W1, X) + b1
A1 = np.tanh(Z1)
Z2 = np.dot(W2, A1) + b2
A2 = sigmoid(Z2)
cache = {'X': X, 'Z1': Z1, 'A1': A1, 'Z2': Z2, 'A2': A2}
return A2, cache
def __one_layer_forward(self, A_prev, W, b, activation, keep_prob):
Z = np.dot(W, A_prev) + b
if activation == 'sigmoid':
A = sigmoid(Z)
if activation == 'relu':
A = relu(Z)
if activation == 'leaky_relu':
A = leaky_relu(Z)
if activation == 'tanh':
A = np.tanh(Z)
if keep_prob == 1:
D = np.ones((A.shape[0], A.shape[1]))
else:
D = np.random.rand(A.shape[0], A.shape[1])
D = D < keep_prob
A = A * D
A = A / keep_prob
cache = {'Z': Z, 'A': A, 'D': D}
return A, cache
def predict_prob(self, X):
n_samples = X.shape[0]
return sigmoid(np.dot(X, self.W) + self.b).reshape(n_samples,)
def sigmoid_backward(self, dA, Z):
s = sigmoid(Z)
dZ = dA * s * (1 - s)
return dZ
def __forward_propagation(self, X):
m = X.shape[1]
A = sigmoid(np.dot(self.w.T, X) + self.b)
return A