反向传播算法的理解

import numpy as np

def nonlin(x, deriv = False):
    if(deriv == True):
        return x * (1 - x)
    return 1 / (1 + np.exp(-x))
//计算s函数，以及用于反向传播的函数


X = np.array([[0.35], [0.9]])
y = np.array([[0.5]])

np.random.seed(1)

W0 = np.array([[0.1, 0.8], [0.4, 0.6]])
W1 = np.array([[0.3, 0.9]])

print ‘original ‘, W0, ‘\n‘, W1

for j in xrange(100):
    l0 = X
    l1 = nonlin(np.dot(W0, l0))
    l2 = nonlin(np.dot(W1, l1))
    l2_error = y - l2
    Error = 1 / 2.0 * (y-l2)**2
    print ‘Error:‘, Error

    l2_delta = l2_error * nonlin(l2, deriv=True)

    l1_error = l2_delta * W1 #back propagation，误差向前传播
    l1_delta = l1_error * nonlin(l1, deriv=True) //类似l2_delta的计算

    W1 += l2_delta * l1.T
    //采用的是另一种反向传播算法，与用误差和参数梯度计算所得结果一样。通过s函数实现。更新的是输出层的参数，仅需一次运算
    W0 += l0.T.dot(l1_delta)
    print W0, ‘\n‘, W1