用Python实现卷积神经网络(二)

我们用python代码构建CNN

1. 导入相关的包:

# import relevant packages

from keras.datasets import mnist import matplotlib.pyplot as plt

%matplotlib   inline import numpy as np

from keras.datasets import mnist from keras.models import Sequential from keras.layers import Dense

from keras.layers import Dropout from keras.utils import np_utils from  keras.layers  import  Flatten

from   keras.layers.convolutional   import   Conv2D

from keras.layers.convolutional import MaxPooling2D from  keras.utils  import  np_utils

from keras import backend as K from  keras  import  regularizers

1. 创建简单的数据集:

# Create a simple dataset X_train=np.array([[[1,2,3,4],[2,3,4,5],[5,6,7,8],[1,3,4,5]],

[[-1,2,3,-4],[2,-3,4,5],[-5,6,-7,8],[-1,-3,-4,-5]]])

y_train=np.array([0,1])

1. 归一化输入:

X_train = X_train / 8

1. 独热编码输出:

y_train    =    np_utils.to_categorical(y_train)

1. 一旦有了只有两个 4 × 4大小的输入和两个输出的数据集,我们首先改变输入形状到要求的格式 (即:样本数,图像高, 图像宽,图像通道数):

X_train = X_train.reshape(X_train.shape[0],X_train.shape[1], X_train.shape[1],1     ).astype('float32')

1. 模建模型:

model   =   Sequential()

model.add(Conv2D(1, (3,3), input_shape=(4,4,1), activation='relu')) model.add(MaxPooling2D(pool_size=(2,  2))) model.add(Flatten())

model.add(Dense(10,  activation='relu')) model.add(Dense(2,   activation='softmax'))

model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])

model.summary()

1. 拟合模型:

model.fit(X_train,  y_train,  epochs=100,  batch_size=2,  verbose=1)

上面模型的不同层如下:

model.layers

不同层的名字和形状如下:

names = [weight.name for layer in model.layers for weight in layer.weights] weights = model.get_weights()

打印zip(names, weights)的名字和权重: print(name,  weight.shape)

某个层的权重用如下方法提取:

model.layers[0].get_weights()

第一个输入的预测计算如下:

model.predict(X_train[0].reshape(1,4,4,1))

现在我们知道对于数字0的概率为0.89066。