Deep Learning菜鸡篇，我的第一个深度神经网络
2017-11-18 17:37 张小贤TT 阅读(587) 评论(0) 编辑 收藏 举报看了一大堆一大堆的框架，作为一个low B , 我还是喜欢从底层实现开始，看了吴恩达Coursera上的视频，同时也在网站上做了一些编程练习，不得不说课程的质量和练习题的质量都是
杠杠的，很到位，这篇博客算是对第一个课程的总结。
说是深度的神经网络，也就是层数深一点的全连接网络，只是练练手顺便复习一下课程，我用的样本是自己做的，（30000,400）的训练样本，附件里会上传，正样本是car，200个正样本，
200个负样本，测试集大小（30000,100），样本的数量神马的选取的都不合理，勿喷，下面上代码
++++++++++++++++++++++++++这一部分是需要的函数 relu 和 sigmoid
#!/usr/bin/env python
# -*- coding:utf-8 -*-
import numpy as np
def sigmoid(Z):
"""
Implements the sigmoid activation in numpy
Arguments:
Z -- numpy array of any shape
Returns:
A -- output of sigmoid(z), same shape as Z
cache -- returns Z as well, useful during backpropagation
"""
A = 1 / (1 + np.exp(-Z))
cache = Z
return A, cache
def relu(Z):
"""
Implement the RELU function.
Arguments:
Z -- Output of the linear layer, of any shape
Returns:
A -- Post-activation parameter, of the same shape as Z
cache -- a python dictionary containing "A" ; stored for computing the backward pass efficiently
"""
A = np.maximum(0, Z)
assert (A.shape == Z.shape)
cache = Z
return A, cache
# def sigmoid_backprob(dA, cache):
# """
# :param dA:
# :param cache:
# :return:
# """
# Z = cache
# tmp , _ = sigmoid(Z)
# dZ = dA * tmp * (1 - tmp)
# return dZ
def sigmoid_backprob(dA, cache):
"""
Implement the backward propagation for a single SIGMOID unit.
Arguments:
dA -- post-activation gradient, of any shape
cache -- 'Z' where we store for computing backward propagation efficiently
Returns:
dZ -- Gradient of the cost with respect to Z
"""
Z = cache
s = 1/(1+np.exp(-Z))
dZ = dA * s * (1-s)
assert (dZ.shape == Z.shape)
return dZ
def relu_backprob(dA, cache):
Z = cache
Daz = Z > 0
dZ = dA * Daz
return dZ
# def relu_backprob(dA, cache):
# """
# Implement the backward propagation for a single RELU unit.
#
# Arguments:
# dA -- post-activation gradient, of any shape
# cache -- 'Z' where we store for computing backward propagation efficiently
#
# Returns:
# dZ -- Gradient of the cost with respect to Z
# """
#
# Z = cache.reshape(dA.shape)
# dZ = np.array(dA, copy=True) # just converting dz to a correct object.
#
# # When z <= 0, you should set dz to 0 as well.
# dZ[Z <= 0] = 0
#
# assert (dZ.shape == Z.shape)
#
# return dZ
++++++++++++++++++++++++++这一部分是样本的读取++++++++++++++++++++++++++++++++++++++
#!/usr/bin/env python
# -*- coding:utf-8 -*-
import numpy as np
def load_data_set():
train_set = np.load('train_set.npy')
train_label = np.load('train_label.npy')
test_set = np.load('test_set.npy')
test_label = np.load('test_label.npy')
return train_set, train_label, test_set, test_label
+++++++++++++++++++++++网络+++++++++++++++++++++++++++++
#!/usr/bin/env python
# -*- coding:utf-8 -*-
import numpy as np
from activation_function import sigmoid, relu, sigmoid_backprob, relu_backprob
from read_data import load_data_set
train_set,train_label,test_set, test_label = load_data_set()
def init_parameters_deep(layer_items):
paramters= {}
L= len(layer_items)
for l in range(1,L):
W = np.random.randn(layer_items[l], layer_items[l-1])*0.01
b = np.zeros((layer_items[l], 1))
paramters['W'+str(l)]= W
paramters['b'+str(l)]= b
return paramters
def linear_forward(A, W, b):
Z = np.dot(W, A) + b
cache = (A, W, b)
return Z, cache
def linear_activation_forward(A_prev, W, b, activation):
Z, linear_cache = linear_forward(A_prev, W, b)
if activation == 'sigmoid':
A, activation_cache = sigmoid(Z)
elif activation == 'relu':
A, activation_cache = relu(Z)
cache = (linear_cache, activation_cache)
return A, cache
def L_forward_model(X,parameters):
caches = []
L = len(parameters) // 2
A = X
for l in range(1, L):
A_prev = A
A, cache = linear_activation_forward(A_prev, parameters['W'+str(l)], parameters['b'+str(l)], activation= 'relu')
caches.append(cache)
AL, cache = linear_activation_forward(A, parameters['W'+str(L)], parameters['b'+str(L)], activation= 'sigmoid')
caches.append(cache)
return AL, caches
def compute_cost(AL, Y,):
m = AL.shape[1]
cost = -1/m *(Y* np.log(AL) + (1 - Y)* np.log(1- AL)).sum()
cost = np.squeeze(cost)
assert (cost.shape == ())
return cost
def linear_back(dZ, cache):
A_prev, W, b = cache
m = A_prev.shape[1]
dW = 1/m * np.dot(dZ, A_prev.T)
db = 1/m * np.sum(dZ, axis=1, keepdims=True)
dA_prev = np.dot(W.T, dZ)
return dA_prev, dW, db
def linear_activation_back(dA, cache, activation):
linear_cache, activation_cache = cache
if activation == 'sigmoid':
dZ = sigmoid_backprob(dA, activation_cache)
dA_prev, dW, db = linear_back(dZ, linear_cache)
elif activation == 'relu':
dZ = relu_backprob(dA, activation_cache)
dA_prev, dW, db = linear_back(dZ, linear_cache)
return dA_prev, dW, db
def L_backprob_model(AL, Y, caches):
grads = {}
dAL = - (np.divide(Y, AL) - np.divide(1 - Y, 1 - AL))
L = len(caches)
cache = caches[L-1]
grads['dA'+str(L-1)] , grads['dW'+str(L)], grads['db'+str(L)] = linear_activation_back(dAL, cache, activation= 'sigmoid')
for l in reversed(range(L- 1)):
cache = caches[l]
grads['dA' + str(l)], grads['dW' + str(l+1)], grads['db' + str(l+1)] = linear_activation_back(dAL, cache,activation='relu')
return grads
def update_parameters(parameters, grads, learning_rate = 0.01):
L = len(parameters) // 2
for l in range(1,L+1):
parameters['W' + str(l)] = parameters['W' + str(l)] - learning_rate * grads['dW' + str(l)]
parameters['b' + str(l)] = parameters['b' + str(l)] - learning_rate * grads['db' + str(l)]
return parameters
def L_nn_deep(layer_items, X, Y, num_iter, learning_rate,print_cost = False):
costs = []
parameters = init_parameters_deep(layer_items)
for i in range(num_iter):
AL, caches = L_forward_model(X, parameters)
cost = compute_cost(AL, Y)
grads = L_backprob_model(AL, Y, caches)
parameters = update_parameters(parameters ,grads, learning_rate)
if i%100 == 0:
costs.append(cost)
if print_cost:
print('第%d次迭代，cost:%f'%(i, cost))
return parameters
layers = [30000, 10000, 5000, 1000, 100,1]
p = L_nn_deep(layers, train_set, train_label, 2000, 0.01, True)
我这个参数神马的，还有数据集都坑的要死，反正我的内存被直接刷爆了，，懒得改了
