当编写完一个深层的网络时,可能求导方式过于复杂稍微不小心就会出错,在开始训练使用这个网络模型之前我们可以先进行梯度检查。
梯度检查的步骤如下:
然后反向传播计算loss的导数grad,用以下公式计算误差:
通常来说,当
ϵ
\epsilon
ϵ为
1
0
−
7
10^{-7}
10−7时,误差在
1
0
−
7
10^{-7}
10−7数量级或者小于
1
0
−
7
10^{-7}
10−7,基本上就没有错误。
下面开始代码部分(假设3层网络)。
首先写入参数格式转换所需的一些函数gc_utils.py:
# -*- coding: utf-8 -*-
import numpy as np
import matplotlib.pyplot as plt
def sigmoid(x):
"""
Compute the sigmoid of x
Arguments:
x -- A scalar or numpy array of any size.
Return:
s -- sigmoid(x)
"""
s = 1/(1+np.exp(-x))
return s
def relu(x):
"""
Compute the relu of x
Arguments:
x -- A scalar or numpy array of any size.
Return:
s -- relu(x)
"""
s = np.maximum(0,x)
return s
def dictionary_to_vector(parameters):
"""
Roll all our parameters dictionary into a single vector satisfying our specific required shape.
"""
keys = []
count = 0
for key in ["W1", "b1", "W2", "b2", "W3", "b3"]:
# flatten parameter
new_vector = np.reshape(parameters[key], (-1,1)) # 将元素转化为一行(列值为1)
keys = keys + [key]*new_vector.shape[0]
if count == 0:
theta = new_vector
else:
theta = np.concatenate((theta, new_vector), axis=0)
count = count + 1
return theta, keys
def vector_to_dictionary(theta):
"""
Unroll all our parameters dictionary from a single vector satisfying our specific required shape.
"""
parameters = {}
parameters["W1"] = theta[:20].reshape((5,4))
parameters["b1"] = theta[20:25].reshape((5,1))
parameters["W2"] = theta[25:40].reshape((3,5))
parameters["b2"] = theta[40:43].reshape((3,1))
parameters["W3"] = theta[43:46].reshape((1,3))
parameters["b3"] = theta[46:47].reshape((1,1))
return parameters
def gradients_to_vector(gradients):
"""
Roll all our gradients dictionary into a single vector satisfying our specific required shape.
"""
count = 0
for key in ["dW1", "db1", "dW2", "db2", "dW3", "db3"]:
# flatten parameter
new_vector = np.reshape(gradients[key], (-1,1))
if count == 0:
theta = new_vector
else:
theta = np.concatenate((theta, new_vector), axis=0)
count = count + 1
return theta
函数dictionary_to_vector()将"parameters" 字典转换为一个称为 “values"的向量,通过将所有参数(W1,b1,W2,b2,W3,b3)reshape为列向量并将它们连接起来而获得。反函数是”vector_to_dictionary",它返回“parameters”字典用于正向传播求loss。
以下为测试代码:
先添加等会测试使用的例子。
import numpy as np
import gc_utils
def gradient_check_n_test_case():
np.random.seed(1)
x = np.random.randn(4, 3)
y = np.array([1, 1, 0])
W1 = np.random.randn(5, 4)
b1 = np.random.randn(5, 1)
W2 = np.random.randn(3, 5)
b2 = np.random.randn(3, 1)
W3 = np.random.randn(1, 3)
b3 = np.random.randn(1, 1)
parameters = {"W1": W1,
"b1": b1,
"W2": W2,
"b2": b2,
"W3": W3,
"b3": b3}
return x, y, parameters
然后是前向传播与反向传播:
def forward_propagation_n(X, Y, parameters):
"""
实现图中的前向传播(并计算成本)。
参数:
X - 训练集为m个例子
Y - m个示例的标签
parameters - 包含参数“W1”,“b1”,“W2”,“b2”,“W3”,“b3”的python字典:
W1 - 权重矩阵,维度为(5,4)
b1 - 偏向量,维度为(5,1)
W2 - 权重矩阵,维度为(3,5)
b2 - 偏向量,维度为(3,1)
W3 - 权重矩阵,维度为(1,3)
b3 - 偏向量,维度为(1,1)
返回:
cost - 成本函数(logistic)
"""
m = X.shape[1]
W1 = parameters["W1"]
b1 = parameters["b1"]
W2 = parameters["W2"]
b2 = parameters["b2"]
W3 = parameters["W3"]
b3 = parameters["b3"]
# LINEAR -> RELU -> LINEAR -> RELU -> LINEAR -> SIGMOID
Z1 = np.dot(W1, X) + b1
A1 = gc_utils.relu(Z1)
Z2 = np.dot(W2, A1) + b2
A2 = gc_utils.relu(Z2)
Z3 = np.dot(W3, A2) + b3
A3 = gc_utils.sigmoid(Z3)
# 计算成本
logprobs = np.multiply(-np.log(A3), Y) + np.multiply(-np.log(1 - A3), 1 - Y)
cost = (1 / m) * np.sum(logprobs)
cache = (Z1, A1, W1, b1, Z2, A2, W2, b2, Z3, A3, W3, b3)
return cost, cache
def backward_propagation_n(X, Y, cache):
"""
实现图中所示的反向传播。
参数:
X - 输入数据点(输入节点数量,1)
Y - 标签
cache - 来自forward_propagation_n()的cache输出
返回:
gradients - 一个字典,其中包含与每个参数、激活和激活前变量相关的成本梯度。
"""
m = X.shape[1]
(Z1, A1, W1, b1, Z2, A2, W2, b2, Z3, A3, W3, b3) = cache
dZ3 = A3 - Y
dW3 = 1. / m * np.dot(dZ3, A2.T)
db3 = 1. / m * np.sum(dZ3, axis=1, keepdims=True)
dA2 = np.dot(W3.T, dZ3)
dZ2 = np.multiply(dA2, np.int64(A2 > 0))
dW2 = 1. / m * np.dot(dZ2, A1.T) * 2 # Should not multiply by 2
# dW2 = 1. / m * np.dot(dZ2, A1.T)
db2 = 1. / m * np.sum(dZ2, axis=1, keepdims=True)
dA1 = np.dot(W2.T, dZ2)
dZ1 = np.multiply(dA1, np.int64(A1 > 0))
dW1 = 1. / m * np.dot(dZ1, X.T)
db1 = 4. / m * np.sum(dZ1, axis=1, keepdims=True) # Should not multiply by 4
# db1 = 1. / m * np.sum(dZ1, axis=1, keepdims=True)
gradients = {"dZ3": dZ3, "dW3": dW3, "db3": db3,
"dA2": dA2, "dZ2": dZ2, "dW2": dW2, "db2": db2,
"dA1": dA1, "dZ1": dZ1, "dW1": dW1, "db1": db1}
return gradients
然后开始梯度检验:
这里大概的逻辑是遍历每个参数用上面的公式求得每个参数的grandapprox,然后总的grandapprox向量进行求误差的计算。
def gradient_check_n(parameters, gradients, X, Y, epsilon=1e-7):
"""
检查backward_propagation_n是否正确计算forward_propagation_n输出的成本梯度
参数:
parameters - 包含参数“W1”,“b1”,“W2”,“b2”,“W3”,“b3”的python字典:
grad_output_propagation_n的输出包含与参数相关的成本梯度。
x - 输入数据点,维度为(输入节点数量,1)
y - 标签
epsilon - 计算输入的微小偏移以计算近似梯度
返回:
difference - 近似梯度和后向传播梯度之间的差异
"""
# 初始化参数
parameters_values, keys = gc_utils.dictionary_to_vector(parameters) # keys用不到
grad = gc_utils.gradients_to_vector(gradients)
num_parameters = parameters_values.shape[0]
J_plus = np.zeros((num_parameters, 1))
J_minus = np.zeros((num_parameters, 1))
gradapprox = np.zeros((num_parameters, 1))
# 计算gradapprox
for i in range(num_parameters):
# 计算J_plus [i]。输入:“parameters_values,epsilon”。输出=“J_plus [i]”
thetaplus = np.copy(parameters_values) # Step 1
thetaplus[i][0] = thetaplus[i][0] + epsilon # Step 2
J_plus[i], cache = forward_propagation_n(X, Y, gc_utils.vector_to_dictionary(thetaplus)) # Step 3 ,cache用不到
# 计算J_minus [i]。输入:“parameters_values,epsilon”。输出=“J_minus [i]”。
thetaminus = np.copy(parameters_values) # Step 1
thetaminus[i][0] = thetaminus[i][0] - epsilon # Step 2
J_minus[i], cache = forward_propagation_n(X, Y, gc_utils.vector_to_dictionary(thetaminus)) # Step 3 ,cache用不到
# 计算gradapprox[i]
gradapprox[i] = (J_plus[i] - J_minus[i]) / (2 * epsilon)
# 通过计算差异比较gradapprox和后向传播梯度。
numerator = np.linalg.norm(grad - gradapprox) # Step 1'
denominator = np.linalg.norm(grad) + np.linalg.norm(gradapprox) # Step 2'
difference = numerator / denominator # Step 3'
if difference < 1e-6:
print("梯度检查:梯度正常!")
else:
print("梯度检查:梯度超出阈值!")
print(difference)
return difference
X, Y, parameters = gradient_check_n_test_case() # 自定义的简易数据集
cost, cache = forward_propagation_n(X, Y, parameters)
gradients = backward_propagation_n(X, Y, cache)
difference = gradient_check_n(parameters, gradients, X, Y)
运行后结果如下:
梯度检查:梯度超出阈值!
0.2850931566540251
显然反向传播出现了问题,我们进行检查,发现是dW2和db1出现问题,进行修改后再次运行:
梯度检查:梯度正常!
1.1885552035482147e-07
注意,如果网络很深参数量势必会很大,计算时间会很长,所以一般训练会关闭梯度检查,在训练之前先进行检查,没问题后进行训练。