卷积神经网络CNN

全连接网络的局限性

• 合理的结构减少参数

卷积神经网络

CNN是一个多层的神经网络，每层由多个二维平面组成， 其中每个平面由多个独立神经元组成。

• CNN结构与全连接神经网络的异同:
  • 比全连接神经网络多了 前面的卷积层跟降采样层次

卷积

注：红色权值矩阵即为卷积核
卷积动态图演示：https://blog.csdn.net/LN_IOS/article/details/80538570

---

池化（降采样）

池化操作跟卷积核基本一致（不再是求点积 而是求最大值或者中位数）

Tensorflow中CNN的相关函数

CIFAR-10图像识别

• 显示数据
  
  
  
  
  

完整代码

#数据加载
%matplotlib inline
import matplotlib.pyplot as plt
import tensorflow as tf
import numpy as np

def load_batch(name):
    with open(name,'rb') as f:
        data_dict = np.load(f,encoding='bytes')
        images = data_dict[b'data']
        labels = data_dict[b'labels']

        images = images.reshape(10000,3,32,32)
        images = images.transpose(0,2,3,1)

        labels = np.array(labels)
        return images,labels

def load_data():
    images_train = []
    labels_train = []
    
    for i in range(5):
        f = 'data_batch_%d'%(i+1)
        image_batch,label_batch=load_batch(f)
        images_train.append(image_batch)
        labels_train.append(label_batch)
        
        Xtrain = np.concatenate(images_train)
        Ytrain = np.concatenate(labels_train)
        del image_batch,label_batch
        
    Xtest,Ytest = load_batch('test_batch')
    print('finished loadding~~~')
    
    return Xtrain,Ytrain,Xtest,Ytest

Xtrain,Ytrain,Xtest,Ytest = load_data()

plt.imshow(Xtrain[6])

label_dict = {0:"airplane",1:"automobile",2:"bird",3:"cat",4:"deer",5:"dog",6:"frog",7:"horse",8:"ship",9:"trunk"}

def plot_images_labels_prediction(images,labels,prediction,idx,num=10):
    fig = plt.gcf()
    fig.set_size_inches(12,6)
    if num > 10:
        num = 10
    for i in range(0,num):
        ax = plt.subplot(2,5,1+i)
        ax.imshow(images[idx],cmap="binary")
        
        title = str(i)+','+label_dict[labels[idx]]
        if len(prediction)>0:
            title+='=>'+label_dict[prediction[idx]]
        ax.set_title(title,fontsize=10)
        
        idx += 1
    plt.show()
    
plot_images_labels_prediction(Xtest,Ytest,[],1,10)

#数据预处理
# 独热编码
from sklearn.preprocessing import OneHotEncoder
encoder = OneHotEncoder(sparse=False)

yy = [[0],[1],[2],[3],[4],[5],[6],[7],[8],[9]]
encoder.fit(yy)
Ytrain_reshape =Ytrain.reshape(-1,1)
Ytrain_onehot = encoder.transform(Ytrain_reshape)
Ytest_reshape = Ytest.reshape(-1,1)
Ytest_onehot = encoder.transform(Ytest_reshape)

Ytrain[:10]

Ytrain_onehot.shape

Ytrain[:5]

Ytrain_onehot[:5]

#定义共享函数
#定义权值
def weight(shape):
    return tf.Variable(tf.truncated_normal(shape,stddev=0.1),name='W')

#定义偏值
def bias(shape):
    return tf.Variable(tf.constant(0.1,shape=shape),name='b')

#定义卷积操作
#步长为2，即原尺寸长和宽各除2
def max_pool_2x2(x):
    return tf.nn.max_pool(x,ksize=[1,2,2,1],strides=[1,2,2,1],padding='SAME')