参考:https://blog.csdn.net/u013733326/article/details/79767169

希望大家直接到上面的网址去查看代码,下面是本人的笔记

实现多层神经网络

1.准备软件包

import numpy as np
import h5py
import matplotlib.pyplot as plt
import testCases #参见资料包,或者在文章底部copy
from dnn_utils import sigmoid, sigmoid_backward, relu, relu_backward #参见资料包
import lr_utils #参见资料包,或者在文章底部copy

为了和作者的数据匹配,需要指定随机种子

np.random.seed(1)

2.初始化参数

def initialize_parameters_deep(layers_dims):
"""
此函数是为了初始化多层网络参数而使用的函数。
参数:
layers_dims - 包含我们网络中每个图层的节点数量的列表 返回:
parameters - 包含参数“W1”,“b1”,...,“WL”,“bL”的字典:
W1 - 权重矩阵,维度为(layers_dims [],layers_dims [-])
bl - 偏向量,维度为(layers_dims [],)
"""
np.random.seed()
parameters = {}
L = len(layers_dims) for l in range(,L):
parameters["W" + str(l)] = np.random.randn(layers_dims[l], layers_dims[l - ]) / np.sqrt(layers_dims[l - ])
parameters["b" + str(l)] = np.zeros((layers_dims[l], )) #确保我要的数据的格式是正确的
assert(parameters["W" + str(l)].shape == (layers_dims[l], layers_dims[l-]))
assert(parameters["b" + str(l)].shape == (layers_dims[l], )) return parameters

测试两层时:

#测试initialize_parameters_deep
print("==============测试initialize_parameters_deep==============")
layers_dims = [,,] #这个其实也是实现了两层
parameters = initialize_parameters_deep(layers_dims)
print(parameters)
print("W1 = " + str(parameters["W1"]))
print("b1 = " + str(parameters["b1"]))
print("W2 = " + str(parameters["W2"]))
print("b2 = " + str(parameters["b2"]))

返回:

==============测试initialize_parameters_deep==============
{'W1': array([[ 0.79989897, 0.19521314, 0.04315498, -0.83337927, -0.12405178],
[-0.15865304, -0.03700312, -0.28040323, -0.01959608, -0.21341839],
[-0.58757818, 0.39561516, 0.39413741, 0.76454432, 0.02237573],
[-0.18097724, -0.24389238, -0.69160568, 0.43932807, -0.49241241]]), 'b1': array([[.],
[.],
[.],
[.]]), 'W2': array([[-0.59252326, -0.10282495, 0.74307418, 0.11835813],
[-0.51189257, -0.3564966 , 0.31262248, -0.08025668],
[-0.38441818, -0.11501536, 0.37252813, 0.98805539]]), 'b2': array([[.],
[.],
[.]])}
W1 = [[ 0.79989897 0.19521314 0.04315498 -0.83337927 -0.12405178]
[-0.15865304 -0.03700312 -0.28040323 -0.01959608 -0.21341839]
[-0.58757818 0.39561516 0.39413741 0.76454432 0.02237573]
[-0.18097724 -0.24389238 -0.69160568 0.43932807 -0.49241241]]
b1 = [[.]
[.]
[.]
[.]]
W2 = [[-0.59252326 -0.10282495 0.74307418 0.11835813]
[-0.51189257 -0.3564966 0.31262248 -0.08025668]
[-0.38441818 -0.11501536 0.37252813 0.98805539]]
b2 = [[.]
[.]
[.]]

测试三层时:

#测试initialize_parameters_deep
print("==============测试initialize_parameters_deep==============")
layers_dims = [,,,] #实现三层看看
parameters = initialize_parameters_deep(layers_dims)
print(parameters)
print("W1 = " + str(parameters["W1"]))
print("b1 = " + str(parameters["b1"]))
print("W2 = " + str(parameters["W2"]))
print("b2 = " + str(parameters["b2"]))
print("W3 = " + str(parameters["W3"]))
print("b3 = " + str(parameters["b3"]))

返回:

==============测试initialize_parameters_deep==============
{'W1': array([[ 0.79989897, 0.19521314, 0.04315498, -0.83337927, -0.12405178],
[-0.15865304, -0.03700312, -0.28040323, -0.01959608, -0.21341839],
[-0.58757818, 0.39561516, 0.39413741, 0.76454432, 0.02237573],
[-0.18097724, -0.24389238, -0.69160568, 0.43932807, -0.49241241]]), 'b1': array([[.],
[.],
[.],
[.]]), 'W2': array([[-0.59252326, -0.10282495, 0.74307418, 0.11835813],
[-0.51189257, -0.3564966 , 0.31262248, -0.08025668],
[-0.38441818, -0.11501536, 0.37252813, 0.98805539]]), 'b2': array([[.],
[.],
[.]]), 'W3': array([[-0.71829494, -0.36166197, -0.46405457],
[-1.39665832, -0.53335157, -0.59113495]]), 'b3': array([[.],
[.]])}
W1 = [[ 0.79989897 0.19521314 0.04315498 -0.83337927 -0.12405178]
[-0.15865304 -0.03700312 -0.28040323 -0.01959608 -0.21341839]
[-0.58757818 0.39561516 0.39413741 0.76454432 0.02237573]
[-0.18097724 -0.24389238 -0.69160568 0.43932807 -0.49241241]]
b1 = [[.]
[.]
[.]
[.]]
W2 = [[-0.59252326 -0.10282495 0.74307418 0.11835813]
[-0.51189257 -0.3564966 0.31262248 -0.08025668]
[-0.38441818 -0.11501536 0.37252813 0.98805539]]
b2 = [[.]
[.]
[.]]
W3 = [[-0.71829494 -0.36166197 -0.46405457]
[-1.39665832 -0.53335157 -0.59113495]]
b3 = [[.]
[.]]

3)前向传播

def L_model_forward(X,parameters):
"""
实现[LINEAR-> RELU] *(L-) - > LINEAR-> SIGMOID计算前向传播,也就是多层网络的前向传播,为后面每一层都执行LINEAR和ACTIVATION 参数:
X - 数据,numpy数组,维度为(输入节点数量,示例数)
parameters - initialize_parameters_deep()的输出 返回:
AL - 最后的激活值
caches - 包含以下内容的缓存列表:
linear_relu_forward()的每个cache(有L-1个,索引为从0到L-)
linear_sigmoid_forward()的cache(只有一个,索引为L-)
"""
caches = []
A = X
L = len(parameters) //
for l in range(,L):
A_prev = A
A, cache = linear_activation_forward(A_prev, parameters['W' + str(l)], parameters['b' + str(l)], "relu")
caches.append(cache) #最后一层使用sigmoid函数进行二分类
AL, cache = linear_activation_forward(A, parameters['W' + str(L)], parameters['b' + str(L)], "sigmoid")
caches.append(cache) assert(AL.shape == (,X.shape[])) return AL,caches

上面函数使用的线性激活函数linear_activation_forward:

def linear_activation_forward(A_prev,W,b,activation):
"""
实现LINEAR-> ACTIVATION 这一层的前向传播 参数:
A_prev - 来自上一层(或输入层)的激活,维度为(上一层的节点数量,示例数)
W - 权重矩阵,numpy数组,维度为(当前层的节点数量,前一层的大小)
b - 偏向量,numpy阵列,维度为(当前层的节点数量,)
activation - 选择在此层中使用的激活函数名,字符串类型,【"sigmoid" | "relu"】 返回:
A - 激活函数的输出,也称为激活后的值
cache - 一个包含“linear_cache”和“activation_cache”的字典,我们需要存储它以有效地计算后向传递
""" if activation == "sigmoid":
Z, linear_cache = linear_forward(A_prev, W, b)
A, activation_cache = sigmoid(Z)
elif activation == "relu":
Z, linear_cache = linear_forward(A_prev, W, b)
A, activation_cache = relu(Z) assert(A.shape == (W.shape[],A_prev.shape[]))
cache = (linear_cache,activation_cache) return A,cache

测试函数L_model_forward_test_case():

def L_model_forward_test_case(): #两层
"""
X = np.array([[-1.02387576, 1.12397796],
[-1.62328545, 0.64667545],
[-1.74314104, -0.59664964]])
parameters = {'W1': np.array([[ 1.62434536, -0.61175641, -0.52817175],
[-1.07296862, 0.86540763, -2.3015387 ]]),
'W2': np.array([[ 1.74481176, -0.7612069 ]]),
'b1': np.array([[ .],
[ .]]),
'b2': np.array([[ .]])}
"""
np.random.seed()
X = np.random.randn(,)
W1 = np.random.randn(,)
b1 = np.random.randn(,)
W2 = np.random.randn(,)
b2 = np.random.randn(,)
parameters = {"W1": W1,
"b1": b1,
"W2": W2,
"b2": b2} return X, parameters

测试:

#测试L_model_forward
print("==============测试L_model_forward==============")
X,parameters = testCases.L_model_forward_test_case()
print(parameters)
AL,caches = L_model_forward(X,parameters)
print("AL = " + str(AL))
print("caches 的长度为 = " + str(len(caches)))
print(caches)

返回:

==============测试L_model_forward==============
{'W1': array([[ 0.3190391 , -0.24937038, 1.46210794, -2.06014071],
[-0.3224172 , -0.38405435, 1.13376944, -1.09989127],
[-0.17242821, -0.87785842, 0.04221375, 0.58281521]]), 'b1': array([[-1.10061918],
[ 1.14472371],
[ 0.90159072]]), 'W2': array([[ 0.50249434, 0.90085595, -0.68372786]]), 'b2': array([[-0.12289023]])}
AL = [[0.17007265 0.2524272 ]]
caches 的长度为 =
[((array([[ 1.62434536, -0.61175641],
[-0.52817175, -1.07296862],
[ 0.86540763, -2.3015387 ],
[ 1.74481176, -0.7612069 ]]), array([[ 0.3190391 , -0.24937038, 1.46210794, -2.06014071],
[-0.3224172 , -0.38405435, 1.13376944, -1.09989127],
[-0.17242821, -0.87785842, 0.04221375, 0.58281521]]), array([[-1.10061918],
[ 1.14472371],
[ 0.90159072]])), array([[-2.77991749, -2.82513147],
[-0.11407702, -0.01812665],
[ 2.13860272, 1.40818979]])), ((array([[. , . ],
[. , . ],
[2.13860272, 1.40818979]]), array([[ 0.50249434, 0.90085595, -0.68372786]]), array([[-0.12289023]])), array([[-1.58511248, -1.08570881]]))]

4.计算成本

def compute_cost(AL,Y):
"""
实施等式(4)定义的成本函数。 参数:
AL - 与标签预测相对应的概率向量,维度为(1,示例数量)
Y - 标签向量(例如:如果不是猫,则为0,如果是猫则为1),维度为(1,数量) 返回:
cost - 交叉熵成本
"""
m = Y.shape[1]
cost = -np.sum(np.multiply(np.log(AL),Y) + np.multiply(np.log(1 - AL), 1 - Y)) / m cost = np.squeeze(cost)
assert(cost.shape == ()) return cost

测试函数:

def compute_cost_test_case():
Y = np.asarray([[1, 1, 1]])
aL = np.array([[.8,.9,0.4]]) return Y, aL

测试:

#测试compute_cost
print("==============测试compute_cost==============")
Y,AL = testCases.compute_cost_test_case()
print("cost = " + str(compute_cost(AL, Y)))

返回:

==============测试compute_cost==============
cost = 0.414931599615397

5.后向传播

因为最后的输出层使用的是sigmoid函数,隐藏层使用的是Relu函数

所以需要对最后一层进行特殊计算,其他层迭代即可

即A,它属于输出层的输出,A=σ(Z),所以我们需要计算dA,我们可以使用下面的代码来计算它:

dAL = - (np.divide(Y, AL) - np.divide( - Y,  - AL))

计算完了以后,我们可以使用此激活后的梯度dA继续向后计算

其实是先通过线性激活部分后向传播得到dz,然后再将dz带入线性部分的后向传播得到dw,db,dA_prev

1)线性部分

吴恩达课后作业学习1-week4-homework-multi-hidden-layer -2-LMLPHP

根据这三个公式来构建后向传播函数

def linear_backward(dZ,cache):
"""
为单层实现反向传播的线性部分(第L层) 参数:
dZ - 相对于(当前第l层的)线性输出的成本梯度
cache - 来自当前层前向传播的值的元组(A_prev,W,b) 返回:
dA_prev - 相对于激活(前一层l-1)的成本梯度,与A_prev维度相同
dW - 相对于W(当前层l)的成本梯度,与W的维度相同
db - 相对于b(当前层l)的成本梯度,与b维度相同
"""
A_prev, W, b = cache
m = A_prev.shape[1]
dW = np.dot(dZ, A_prev.T) / m
db = np.sum(dZ, axis=1, keepdims=True) / m
dA_prev = np.dot(W.T, dZ) assert (dA_prev.shape == A_prev.shape)
assert (dW.shape == W.shape)
assert (db.shape == b.shape) return dA_prev, dW, db

2)线性激活部分

将线性部分也使用了进来

在dnn_utils.py中定义了两个现成可用的后向函数,用来帮助计算dz:

如果 g(.)是激活函数, 那么sigmoid_backward 和 relu_backward 这样计算:
吴恩达课后作业学习1-week4-homework-multi-hidden-layer -2-LMLPHP

  • sigmoid_backward:实现了sigmoid()函数的反向传播,用来计算dz为:
dZ = sigmoid_backward(dA, activation_cache)
  • relu_backward: 实现了relu()函数的反向传播,用来计算dz为:
dZ = relu_backward(dA, activation_cache)

后向函数为:

def sigmoid_backward(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_backward(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
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

代码为:

def linear_activation_backward(dA,cache,activation="relu"):
"""
实现LINEAR-> ACTIVATION层的后向传播。 参数:
dA - 当前层l的激活后的梯度值
cache - 我们存储的用于有效计算反向传播的值的元组(值为linear_cache,activation_cache)
activation - 要在此层中使用的激活函数名,字符串类型,【"sigmoid" | "relu"】
返回:
dA_prev - 相对于激活(前一层l-1)的成本梯度值,与A_prev维度相同
dW - 相对于W(当前层l)的成本梯度值,与W的维度相同
db - 相对于b(当前层l)的成本梯度值,与b的维度相同
"""
linear_cache, activation_cache = cache
#其实是先通过线性激活部分后向传播得到dz,然后再将dz带入线性部分的后向传播得到dw,db,dA_prev
if activation == "relu":
dZ = relu_backward(dA, activation_cache)
dA_prev, dW, db = linear_backward(dZ, linear_cache)
elif activation == "sigmoid":
dZ = sigmoid_backward(dA, activation_cache)
dA_prev, dW, db = linear_backward(dZ, linear_cache) return dA_prev,dW,db

整合函数,用于多层神经网络:

def L_model_backward(AL,Y,caches):
"""
对[LINEAR-> RELU] *(L-) - > LINEAR - > SIGMOID组执行反向传播,就是多层网络的向后传播 参数:
AL - 概率向量,正向传播输出层的输出(L_model_forward())
Y - 标签向量,真正正确的结果(例如:如果不是猫,则为0,如果是猫则为1),维度为(,数量)
caches - 包含以下内容的cache列表:
linear_activation_forward("relu")的cache,不包含输出层
linear_activation_forward("sigmoid")的cache 返回:
grads - 具有梯度值的字典
grads [“dA”+ str(l)] = ...
grads [“dW”+ str(l)] = ...
grads [“db”+ str(l)] = ...
"""
grads = {}
L = len(caches)
m = AL.shape[] #得到数据量,几张照片
Y = Y.reshape(AL.shape) #保证AL和Y两者格式相同
dAL = - (np.divide(Y, AL) - np.divide( - Y, - AL)) #计算得到dAL current_cache = caches[L-] #用于输出层的cache存储的值
#对输出层进行后向传播
grads["dA" + str(L)], grads["dW" + str(L)], grads["db" + str(L)] = linear_activation_backward(dAL, current_cache, "sigmoid") for l in reversed(range(L-)): #迭代对接下来的隐藏层进行后向传播
current_cache = caches[l]
dA_prev_temp, dW_temp, db_temp = linear_activation_backward(grads["dA" + str(l + )], current_cache, "relu")
grads["dA" + str(l + )] = dA_prev_temp
grads["dW" + str(l + )] = dW_temp
grads["db" + str(l + )] = db_temp return grads

测试函数:

def L_model_backward_test_case(): #计算后向传播的前向传播的值
"""
X = np.random.rand(,)
Y = np.array([[, ]])
parameters = {'W1': np.array([[ 1.78862847, 0.43650985, 0.09649747]]), 'b1': np.array([[ .]])} aL, caches = (np.array([[ 0.60298372, 0.87182628]]), [((np.array([[ 0.20445225, 0.87811744],
[ 0.02738759, 0.67046751],
[ 0.4173048 , 0.55868983]]),
np.array([[ 1.78862847, 0.43650985, 0.09649747]]),
np.array([[ .]])),
np.array([[ 0.41791293, 1.91720367]]))])
"""
np.random.seed()
AL = np.random.randn(, )
Y = np.array([[, ]]) A1 = np.random.randn(,)
W1 = np.random.randn(,)
b1 = np.random.randn(,)
Z1 = np.random.randn(,)
linear_cache_activation_1 = ((A1, W1, b1), Z1) A2 = np.random.randn(,)
W2 = np.random.randn(,)
b2 = np.random.randn(,)
Z2 = np.random.randn(,)
linear_cache_activation_2 = ( (A2, W2, b2), Z2) caches = (linear_cache_activation_1, linear_cache_activation_2) return AL, Y, caches

测试:

#测试L_model_backward
print("==============测试L_model_backward==============")
AL, Y_assess, caches = testCases.L_model_backward_test_case()
grads = L_model_backward(AL, Y_assess, caches)
print ("dW1 = "+ str(grads["dW1"]))
print ("db1 = "+ str(grads["db1"]))
print ("dA1 = "+ str(grads["dA1"]))

返回:

==============测试L_model_backward==============
dW1 = [[0.41010002 0.07807203 0.13798444 0.10502167]
[. . . . ]
[0.05283652 0.01005865 0.01777766 0.0135308 ]]
db1 = [[-0.22007063]
[ . ]
[-0.02835349]]
dA1 = [[ . 0.52257901]
[ . -0.3269206 ]
[ . -0.32070404]
[ . -0.74079187]]

6.更新参数

根据上面后向传播得到的dw,db,dA_prev来更新参数,其中 α 是学习率

吴恩达课后作业学习1-week4-homework-multi-hidden-layer -2-LMLPHP

函数:

def update_parameters(parameters, grads, learning_rate):
"""
使用梯度下降更新参数 参数:
parameters - 包含你的参数的字典,即w和b
grads - 包含梯度值的字典,是L_model_backward的输出 返回:
parameters - 包含更新参数的字典
参数[“W”+ str(l)] = ...
参数[“b”+ str(l)] = ...
"""
L = len(parameters) // 2 #整除2,得到层数
for l in range(L):
parameters["W" + str(l + 1)] = parameters["W" + str(l + 1)] - learning_rate * grads["dW" + str(l + 1)]
parameters["b" + str(l + 1)] = parameters["b" + str(l + 1)] - learning_rate * grads["db" + str(l + 1)] return parameters

测试函数:

def update_parameters_test_case():
"""
parameters = {'W1': np.array([[ 1.78862847, 0.43650985, 0.09649747],
[-1.8634927 , -0.2773882 , -0.35475898],
[-0.08274148, -0.62700068, -0.04381817],
[-0.47721803, -1.31386475, 0.88462238]]),
'W2': np.array([[ 0.88131804, 1.70957306, 0.05003364, -0.40467741],
[-0.54535995, -1.54647732, 0.98236743, -1.10106763],
[-1.18504653, -0.2056499 , 1.48614836, 0.23671627]]),
'W3': np.array([[-1.02378514, -0.7129932 , 0.62524497],
[-0.16051336, -0.76883635, -0.23003072]]),
'b1': np.array([[ 0.],
[ 0.],
[ 0.],
[ 0.]]),
'b2': np.array([[ 0.],
[ 0.],
[ 0.]]),
'b3': np.array([[ 0.],
[ 0.]])}
grads = {'dW1': np.array([[ 0.63070583, 0.66482653, 0.18308507],
[ 0. , 0. , 0. ],
[ 0. , 0. , 0. ],
[ 0. , 0. , 0. ]]),
'dW2': np.array([[ 1.62934255, 0. , 0. , 0. ],
[ 0. , 0. , 0. , 0. ],
[ 0. , 0. , 0. , 0. ]]),
'dW3': np.array([[-1.40260776, 0. , 0. ]]),
'da1': np.array([[ 0.70760786, 0.65063504],
[ 0.17268975, 0.15878569],
[ 0.03817582, 0.03510211]]),
'da2': np.array([[ 0.39561478, 0.36376198],
[ 0.7674101 , 0.70562233],
[ 0.0224596 , 0.02065127],
[-0.18165561, -0.16702967]]),
'da3': np.array([[ 0.44888991, 0.41274769],
[ 0.31261975, 0.28744927],
[-0.27414557, -0.25207283]]),
'db1': 0.75937676204411464,
'db2': 0.86163759922811056,
'db3': -0.84161956022334572}
"""
np.random.seed(2)
W1 = np.random.randn(3,4)
b1 = np.random.randn(3,1)
W2 = np.random.randn(1,3)
b2 = np.random.randn(1,1)
parameters = {"W1": W1,
"b1": b1,
"W2": W2,
"b2": b2}
np.random.seed(3)
dW1 = np.random.randn(3,4)
db1 = np.random.randn(3,1)
dW2 = np.random.randn(1,3)
db2 = np.random.randn(1,1)
grads = {"dW1": dW1,
"db1": db1,
"dW2": dW2,
"db2": db2} return parameters, grads

测试:

#测试update_parameters
print("==============测试update_parameters==============")
parameters, grads = testCases.update_parameters_test_case()
parameters = update_parameters(parameters, grads, 0.1) print ("W1 = "+ str(parameters["W1"]))
print ("b1 = "+ str(parameters["b1"]))
print ("W2 = "+ str(parameters["W2"]))
print ("b2 = "+ str(parameters["b2"]))

返回:

==============测试update_parameters==============
W1 = [[-0.59562069 -0.09991781 -2.14584584 1.82662008]
[-1.76569676 -0.80627147 0.51115557 -1.18258802]
[-1.0535704 -0.86128581 0.68284052 2.20374577]]
b1 = [[-0.04659241]
[-1.28888275]
[ 0.53405496]]
W2 = [[-0.55569196 0.0354055 1.32964895]]
b2 = [[-0.84610769]]

7.整合函数——训练

def L_layer_model(X, Y, layers_dims, learning_rate=0.0075, num_iterations=, print_cost=False,isPlot=True):
"""
实现一个L层神经网络:[LINEAR-> RELU] *(L-) - > LINEAR-> SIGMOID。 参数:
X - 输入的数据,维度为(n_x,例子数)
Y - 标签,向量,0为非猫,1为猫,维度为(,数量)
layers_dims - 层数的向量,维度为(n_y,n_h,···,n_h,n_y)
learning_rate - 学习率
num_iterations - 迭代的次数
print_cost - 是否打印成本值,每100次打印一次
isPlot - 是否绘制出误差值的图谱 返回:
parameters - 模型学习的参数。 然后他们可以用来预测。
"""
np.random.seed()
costs = [] #随机初始化参数
parameters = initialize_parameters_deep(layers_dims) for i in range(,num_iterations):
AL , caches = L_model_forward(X,parameters) #前向传播 cost = compute_cost(AL,Y) #成本计算 grads = L_model_backward(AL,Y,caches) #后向传播 parameters = update_parameters(parameters,grads,learning_rate) #更新参数 #打印成本值,如果print_cost=False则忽略
if i % == :
#记录成本
costs.append(cost)
#是否打印成本值
if print_cost:
print("第", i ,"次迭代,成本值为:" ,np.squeeze(cost))
#迭代完成,根据条件绘制图
if isPlot:
plt.plot(np.squeeze(costs))
plt.ylabel('cost')
plt.xlabel('iterations (per tens)')
plt.title("Learning rate =" + str(learning_rate))
plt.show()
return parameters

我们现在开始加载数据集,图像数据集的处理可以参照吴恩达课后作业学习1-week2-homework-logistic

train_set_x_orig , train_set_y , test_set_x_orig , test_set_y , classes = lr_utils.load_dataset()

train_x_flatten = train_set_x_orig.reshape(train_set_x_orig.shape[], -).T
test_x_flatten = test_set_x_orig.reshape(test_set_x_orig.shape[], -).T train_x = train_x_flatten /
train_y = train_set_y
test_x = test_x_flatten /
test_y = test_set_y

数据集加载完成,开始正式训练:

layers_dims = [, , , , ] #  -layer model
parameters = L_layer_model(train_x, train_y, layers_dims, num_iterations = , print_cost = True,isPlot=True)

返回:

第  次迭代,成本值为: 0.715731513413713
第 次迭代,成本值为: 0.6747377593469114
第 次迭代,成本值为: 0.6603365433622127
第 次迭代,成本值为: 0.6462887802148751
第 次迭代,成本值为: 0.6298131216927773
第 次迭代,成本值为: 0.606005622926534
第 次迭代,成本值为: 0.5690041263975135
第 次迭代,成本值为: 0.519796535043806
第 次迭代,成本值为: 0.46415716786282285
第 次迭代,成本值为: 0.40842030048298916
第 次迭代,成本值为: 0.37315499216069037
第 次迭代,成本值为: 0.30572374573047123
第 次迭代,成本值为: 0.2681015284774084
第 次迭代,成本值为: 0.23872474827672593
第 次迭代,成本值为: 0.20632263257914712
第 次迭代,成本值为: 0.17943886927493544
第 次迭代,成本值为: 0.15798735818801213
第 次迭代,成本值为: 0.14240413012273928
第 次迭代,成本值为: 0.12865165997885833
第 次迭代,成本值为: 0.11244314998155475
第 次迭代,成本值为: 0.08505631034966661
第 次迭代,成本值为: 0.05758391198605767
第 次迭代,成本值为: 0.044567534546938604
第 次迭代,成本值为: 0.03808275166597662
第 次迭代,成本值为: 0.034410749018403006

图示:

吴恩达课后作业学习1-week4-homework-multi-hidden-layer -2-LMLPHP

8.预测

def predict(X, y, parameters):
"""
该函数用于预测L层神经网络的结果,当然也包含两层 参数:
X - 测试集
y - 标签
parameters - 训练模型得到的最优参数 返回:
p - 给定数据集X的预测
""" m = X.shape[1]
n = len(parameters) // 2 # 神经网络的层数
p = np.zeros((1,m)) #根据参数前向传播
probas, caches = L_model_forward(X, parameters) for i in range(0, probas.shape[1]):
if probas[0,i] > 0.5:
p[0,i] = 1
else:
p[0,i] = 0 print("准确度为: " + str(float(np.sum((p == y))/m))) return p

预测函数构建好了我们就开始预测,查看训练集和测试集的准确性:

pred_train = predict(train_x, train_y, parameters) #训练集
pred_test = predict(test_x, test_y, parameters) #测试集

返回:

准确度为: 0.9952153110047847
准确度为: 0.78

可见多层神经网络训练的效果比两层的要更好一些

05-11 09:32