人工智能从入门到精通（18）

循环神经网络

传统机器学习非常依赖于人工提取特征，就会遇到很多瓶颈的问提

import numpy as np
X=[1,2]
state=[0.0,0.0]
x_cell_state=np.asarray([[0.1,0.2],[0.3,0.4]])
w_cell_input=np.asarray([0.5,0.6])
b_cell=np.asarray([0.1,-0.1])
w_output=np.asarray([[1.0],[2.0]])
b_output=0.1
for i in range(len(X)):
    before_activation=np.dot(state,w_cell_input)+X[i]*w_cell_input+b_cell
    state=np.tanh(before_activation)
    final_output=np.dot(state,w_output)+b_output
    print("before activation:",before_activation)
    print("state:",state)
    print("output:",final_output)