chap5 CNN

卷积神经网络（CNN）

问题描述：

利用卷积神经网络，实现对MNIST数据集的分类问题

数据集：

MNIST数据集包括60000张训练图片和10000张测试图片。图片样本的数量已经足够训练一个很复杂的模型（例如 CNN的深层神经网络）。它经常被用来作为一个新的模式识别模型的测试用例。而且它也是一个方便学生和研究者们执行用例的数据集。除此之外，MNIST数据集是一个相对较小的数据集，可以在你的笔记本CPUs上面直接执行

题目要求

Pytorch版本的卷积神经网络需要补齐self.conv1中的nn.Conv2d()和self.conv2()的参数，还需要填写x=x.view()中的内容。
训练精度应该在96%以上。

import os
import torch
import torch.nn as nn
from torch.autograd import Variable
import torch.utils.data as Data
import torchvision
import torch.nn.functional as F
import numpy as np
learning_rate = 1e-4
keep_prob_rate = 0.7 #
max_epoch = 3
BATCH_SIZE = 50

DOWNLOAD_MNIST = False
if not(os.path.exists('./mnist/')) or not os.listdir('./mnist/'):
    # not mnist dir or mnist is empyt dir
    DOWNLOAD_MNIST = True


train_data = torchvision.datasets.MNIST(root='./mnist/',train=True, transform=torchvision.transforms.ToTensor(), download=DOWNLOAD_MNIST,)
train_loader = Data.DataLoader(dataset = train_data ,batch_size= BATCH_SIZE ,shuffle= True)

test_data = torchvision.datasets.MNIST(root = './mnist/',train = False)
test_x = Variable(torch.unsqueeze(test_data.test_data,dim  = 1),volatile = True).type(torch.FloatTensor)[:500]/255.
test_y = test_data.test_labels[:500].numpy()

class CNN(nn.Module):
    def __init__(self):
        super(CNN, self).__init__()
        self.conv1 = nn.Sequential(
            nn.Conv2d( # ???
                # patch 7 * 7 ; 1  in channels ; 32 out channels ; ; stride is 1
                # padding style is same(that means the convolution opration's input and output have the same size)
                in_channels=1,
                out_channels=32,
                kernel_size=7,
                stride=1,
                padding=3,
            ),
            nn.ReLU(),        # activation function
            nn.MaxPool2d(2),  # pooling operation
        )
        self.conv2 = nn.Sequential( # ???
            # line 1 : convolution function, patch 5*5 , 32 in channels ;64 out channels; padding style is same; stride is 1
            # line 2 : choosing your activation funciont
            # line 3 : pooling operation function.
            nn.Conv2d(in_channels=32, out_channels=64, kernel_size=5, padding=2, stride=1),
            nn.ReLU(),
            nn.AvgPool2d(2),
        )
        self.out1 = nn.Linear( 7*7*64 , 1024 , bias= True)   # full connection layer one

        self.dropout = nn.Dropout(keep_prob_rate)
        self.out2 = nn.Linear(1024,10,bias=True)



    def forward(self, x):
        x = self.conv1(x)
        x = self.conv2(x)
        x = x.view(-1, 7*7*64)  # flatten the output of coonv2 to (batch_size ,32 * 7 * 7)    # ???
        out1 = self.out1(x)
        out1 = F.relu(out1)
        out1 = self.dropout(out1)
        out2 = self.out2(out1)
        output = F.softmax(out2)
        return output


def test(cnn):
    global prediction
    y_pre = cnn(test_x)
    _,pre_index= torch.max(y_pre,1)
    pre_index= pre_index.view(-1)
    prediction = pre_index.data.numpy()
    correct  = np.sum(prediction == test_y)
    return correct / 500.0


def train(cnn):
    optimizer = torch.optim.Adam(cnn.parameters(), lr=learning_rate )
    loss_func = nn.CrossEntropyLoss()
    for epoch in range(max_epoch):
        for step, (x_, y_) in enumerate(train_loader):
            x ,y= Variable(x_),Variable(y_)
            output = cnn(x)  
            loss = loss_func(output,y)
            optimizer.zero_grad()
            loss.backward()
            optimizer.step()
            
            if step != 0 and step % 20 ==0:
                print("=" * 10,step,"="*5,"="*5, "test accuracy is ",test(cnn) ,"=" * 10 )

if __name__ == '__main__':
    cnn = CNN()
    train(cnn)

训练结果为：

========== 20 ===== ===== test accuracy is  0.224 ==========
========== 40 ===== ===== test accuracy is  0.362 ==========
========== 60 ===== ===== test accuracy is  0.402 ==========
========== 80 ===== ===== test accuracy is  0.51 ==========
========== 100 ===== ===== test accuracy is  0.608 ==========
========== 120 ===== ===== test accuracy is  0.624 ==========
========== 140 ===== ===== test accuracy is  0.708 ==========
========== 160 ===== ===== test accuracy is  0.684 ==========
========== 180 ===== ===== test accuracy is  0.738 ==========
========== 200 ===== ===== test accuracy is  0.766 ==========
========== 220 ===== ===== test accuracy is  0.778 ==========
========== 240 ===== ===== test accuracy is  0.796 ==========
========== 260 ===== ===== test accuracy is  0.802 ==========
========== 280 ===== ===== test accuracy is  0.81 ==========
========== 300 ===== ===== test accuracy is  0.812 ==========
========== 320 ===== ===== test accuracy is  0.82 ==========
========== 340 ===== ===== test accuracy is  0.848 ==========
========== 360 ===== ===== test accuracy is  0.83 ==========
========== 380 ===== ===== test accuracy is  0.852 ==========
========== 400 ===== ===== test accuracy is  0.852 ==========
========== 420 ===== ===== test accuracy is  0.856 ==========
========== 440 ===== ===== test accuracy is  0.874 ==========
========== 460 ===== ===== test accuracy is  0.85 ==========
========== 480 ===== ===== test accuracy is  0.874 ==========
========== 500 ===== ===== test accuracy is  0.864 ==========
========== 520 ===== ===== test accuracy is  0.858 ==========
========== 540 ===== ===== test accuracy is  0.884 ==========
========== 560 ===== ===== test accuracy is  0.872 ==========
========== 580 ===== ===== test accuracy is  0.9 ==========
========== 600 ===== ===== test accuracy is  0.88 ==========
========== 620 ===== ===== test accuracy is  0.886 ==========
========== 640 ===== ===== test accuracy is  0.882 ==========
========== 660 ===== ===== test accuracy is  0.886 ==========
========== 680 ===== ===== test accuracy is  0.876 ==========
========== 700 ===== ===== test accuracy is  0.882 ==========
========== 720 ===== ===== test accuracy is  0.886 ==========
========== 740 ===== ===== test accuracy is  0.894 ==========
========== 760 ===== ===== test accuracy is  0.894 ==========
========== 780 ===== ===== test accuracy is  0.9 ==========
========== 800 ===== ===== test accuracy is  0.898 ==========
========== 820 ===== ===== test accuracy is  0.912 ==========
========== 840 ===== ===== test accuracy is  0.894 ==========
========== 860 ===== ===== test accuracy is  0.898 ==========
========== 880 ===== ===== test accuracy is  0.888 ==========
========== 900 ===== ===== test accuracy is  0.896 ==========
========== 920 ===== ===== test accuracy is  0.888 ==========
========== 940 ===== ===== test accuracy is  0.91 ==========
========== 960 ===== ===== test accuracy is  0.908 ==========
========== 980 ===== ===== test accuracy is  0.918 ==========
========== 1000 ===== ===== test accuracy is  0.906 ==========
========== 1020 ===== ===== test accuracy is  0.908 ==========
========== 1040 ===== ===== test accuracy is  0.906 ==========
========== 1060 ===== ===== test accuracy is  0.914 ==========
========== 1080 ===== ===== test accuracy is  0.908 ==========
========== 1100 ===== ===== test accuracy is  0.906 ==========
========== 1120 ===== ===== test accuracy is  0.906 ==========
========== 1140 ===== ===== test accuracy is  0.924 ==========
========== 1160 ===== ===== test accuracy is  0.918 ==========
========== 1180 ===== ===== test accuracy is  0.904 ==========
========== 20 ===== ===== test accuracy is  0.924 ==========
========== 40 ===== ===== test accuracy is  0.908 ==========
========== 60 ===== ===== test accuracy is  0.92 ==========
========== 80 ===== ===== test accuracy is  0.91 ==========
========== 100 ===== ===== test accuracy is  0.926 ==========
========== 120 ===== ===== test accuracy is  0.91 ==========
========== 140 ===== ===== test accuracy is  0.922 ==========
========== 160 ===== ===== test accuracy is  0.932 ==========
========== 180 ===== ===== test accuracy is  0.932 ==========
========== 200 ===== ===== test accuracy is  0.93 ==========
========== 220 ===== ===== test accuracy is  0.94 ==========
========== 240 ===== ===== test accuracy is  0.918 ==========
========== 260 ===== ===== test accuracy is  0.934 ==========
========== 280 ===== ===== test accuracy is  0.93 ==========
========== 300 ===== ===== test accuracy is  0.934 ==========
========== 320 ===== ===== test accuracy is  0.934 ==========
========== 340 ===== ===== test accuracy is  0.93 ==========
========== 360 ===== ===== test accuracy is  0.944 ==========
========== 380 ===== ===== test accuracy is  0.938 ==========
========== 400 ===== ===== test accuracy is  0.92 ==========
========== 420 ===== ===== test accuracy is  0.936 ==========
========== 440 ===== ===== test accuracy is  0.948 ==========
========== 460 ===== ===== test accuracy is  0.934 ==========
========== 480 ===== ===== test accuracy is  0.938 ==========
========== 500 ===== ===== test accuracy is  0.916 ==========
========== 520 ===== ===== test accuracy is  0.916 ==========
========== 540 ===== ===== test accuracy is  0.928 ==========
========== 560 ===== ===== test accuracy is  0.936 ==========
========== 580 ===== ===== test accuracy is  0.942 ==========
========== 600 ===== ===== test accuracy is  0.922 ==========
========== 620 ===== ===== test accuracy is  0.94 ==========
========== 640 ===== ===== test accuracy is  0.94 ==========
========== 660 ===== ===== test accuracy is  0.96 ==========
========== 680 ===== ===== test accuracy is  0.938 ==========
========== 700 ===== ===== test accuracy is  0.936 ==========
========== 720 ===== ===== test accuracy is  0.94 ==========
========== 740 ===== ===== test accuracy is  0.946 ==========
========== 760 ===== ===== test accuracy is  0.946 ==========
========== 780 ===== ===== test accuracy is  0.948 ==========
========== 800 ===== ===== test accuracy is  0.95 ==========
========== 820 ===== ===== test accuracy is  0.948 ==========
========== 840 ===== ===== test accuracy is  0.95 ==========
========== 860 ===== ===== test accuracy is  0.94 ==========
========== 880 ===== ===== test accuracy is  0.956 ==========
========== 900 ===== ===== test accuracy is  0.944 ==========
========== 920 ===== ===== test accuracy is  0.948 ==========
========== 940 ===== ===== test accuracy is  0.95 ==========
========== 960 ===== ===== test accuracy is  0.944 ==========
========== 980 ===== ===== test accuracy is  0.94 ==========
========== 1000 ===== ===== test accuracy is  0.946 ==========
========== 1020 ===== ===== test accuracy is  0.952 ==========
========== 1040 ===== ===== test accuracy is  0.952 ==========
========== 1060 ===== ===== test accuracy is  0.944 ==========
========== 1080 ===== ===== test accuracy is  0.956 ==========
========== 1100 ===== ===== test accuracy is  0.96 ==========
========== 1120 ===== ===== test accuracy is  0.948 ==========
========== 1140 ===== ===== test accuracy is  0.942 ==========
========== 1160 ===== ===== test accuracy is  0.948 ==========
========== 1180 ===== ===== test accuracy is  0.944 ==========
========== 20 ===== ===== test accuracy is  0.952 ==========
========== 40 ===== ===== test accuracy is  0.96 ==========
========== 60 ===== ===== test accuracy is  0.948 ==========
========== 80 ===== ===== test accuracy is  0.954 ==========
========== 100 ===== ===== test accuracy is  0.948 ==========
========== 120 ===== ===== test accuracy is  0.948 ==========
========== 140 ===== ===== test accuracy is  0.958 ==========
========== 160 ===== ===== test accuracy is  0.942 ==========
========== 180 ===== ===== test accuracy is  0.948 ==========
========== 200 ===== ===== test accuracy is  0.952 ==========
========== 220 ===== ===== test accuracy is  0.952 ==========
========== 240 ===== ===== test accuracy is  0.95 ==========
========== 260 ===== ===== test accuracy is  0.966 ==========
========== 280 ===== ===== test accuracy is  0.96 ==========
========== 300 ===== ===== test accuracy is  0.956 ==========
========== 320 ===== ===== test accuracy is  0.96 ==========
========== 340 ===== ===== test accuracy is  0.956 ==========
========== 360 ===== ===== test accuracy is  0.956 ==========
========== 380 ===== ===== test accuracy is  0.954 ==========
========== 400 ===== ===== test accuracy is  0.96 ==========
========== 420 ===== ===== test accuracy is  0.966 ==========
========== 440 ===== ===== test accuracy is  0.96 ==========
========== 460 ===== ===== test accuracy is  0.954 ==========
========== 480 ===== ===== test accuracy is  0.968 ==========
========== 500 ===== ===== test accuracy is  0.958 ==========
========== 520 ===== ===== test accuracy is  0.958 ==========
========== 540 ===== ===== test accuracy is  0.962 ==========
========== 560 ===== ===== test accuracy is  0.968 ==========
========== 580 ===== ===== test accuracy is  0.958 ==========
========== 600 ===== ===== test accuracy is  0.952 ==========
========== 620 ===== ===== test accuracy is  0.95 ==========
========== 640 ===== ===== test accuracy is  0.964 ==========
========== 660 ===== ===== test accuracy is  0.962 ==========
========== 680 ===== ===== test accuracy is  0.96 ==========
========== 700 ===== ===== test accuracy is  0.962 ==========
========== 720 ===== ===== test accuracy is  0.964 ==========
========== 740 ===== ===== test accuracy is  0.958 ==========
========== 760 ===== ===== test accuracy is  0.96 ==========
========== 780 ===== ===== test accuracy is  0.972 ==========
========== 800 ===== ===== test accuracy is  0.962 ==========
========== 820 ===== ===== test accuracy is  0.968 ==========
========== 840 ===== ===== test accuracy is  0.964 ==========
========== 860 ===== ===== test accuracy is  0.96 ==========
========== 880 ===== ===== test accuracy is  0.964 ==========
========== 900 ===== ===== test accuracy is  0.96 ==========
========== 920 ===== ===== test accuracy is  0.96 ==========
========== 940 ===== ===== test accuracy is  0.97 ==========
========== 960 ===== ===== test accuracy is  0.956 ==========
========== 980 ===== ===== test accuracy is  0.966 ==========
========== 1000 ===== ===== test accuracy is  0.964 ==========
========== 1020 ===== ===== test accuracy is  0.964 ==========
========== 1040 ===== ===== test accuracy is  0.97 ==========
========== 1060 ===== ===== test accuracy is  0.974 ==========
========== 1080 ===== ===== test accuracy is  0.962 ==========
========== 1100 ===== ===== test accuracy is  0.97 ==========
========== 1120 ===== ===== test accuracy is  0.974 ==========
========== 1140 ===== ===== test accuracy is  0.978 ==========
========== 1160 ===== ===== test accuracy is  0.976 ==========
========== 1180 ===== ===== test accuracy is  0.974 ==========