sigmoid

def sigmoid(X):  # define activation: sigmoid

    output = 1 / (1 + np.exp(-X))

    return output

def sigmoid_gradient(X):

    output = sigmoid(X)*(1-sigmoid(X))

    return output

def softmax(X):  # define activation: softmax

    return np.exp(X) / np.sum(np.exp(X), axis=1, keepdims=True)

def cross_entropy(p, q):

    epsilon = 1e-15

    H = 0

    for i in range(len(p)):

        H += -p[i]*np.log(q[i]+epsilon)

    H = H.sum()/p.shape[0]

    return H

from tensorflow import keras

# 做 One-hot encoding

y = keras.utils.to_categorical(array)