# -*- coding:utf-8 -*-
#@Time : 2020/6/29 22:29
#@Author: ly
import torch
import torch.nn as nn
import torch.utils.data as Data
import torchvision
import matplotlib.pyplot as plt
# Hyper Parameters
EPOCH = 1 # train the training data n times, to save time, we just train 1 epoch
BATCH_SIZE = 50
LR = 0.001 # learning rate
DOWNLOAD_MNIST = False
# Mnist digits dataset
train_data = torchvision.datasets.MNIST(
root='./mnist/',
train=True, # this is training data
transform=torchvision.transforms.ToTensor(), # Converts a PIL.Image or numpy.ndarray to
# torch.FloatTensor of shape (C x H x W) and normalize in the range [0.0, 1.0]
download=DOWNLOAD_MNIST,
)
# plot one example
#print(train_data.data.size()) # (60000, 28, 28)
#print(train_data.targets.size()) # (60000)
# plt.imshow(train_data.data[0].numpy(), cmap='gray')
# plt.title('%i' % train_data.targets[0])
# plt.show()
# Data Loader for easy mini-batch return in training, the image batch shape will be (50, 1, 28, 28)
train_loader = Data.DataLoader(dataset=train_data, batch_size=BATCH_SIZE, shuffle=True)
# pick 2000 samples to speed up testing
test_data = torchvision.datasets.MNIST(root='./mnist/', train=False)
test_x = torch.unsqueeze(test_data.data, dim=1).type(torch.FloatTensor)[:2000]/255. # shape from (2000, 28, 28) to (2000, 1, 28, 28), value in range(0,1)
test_y = test_data.targets[:2000] #取前两千个数据
#print(test_data.data.size()) # (10000, 28, 28)
#print(test_data.targets.size())
class CNN(nn.Module):
def __init__(self):
super(CNN, self).__init__()
#第一层网络
self.conv1 = nn.Sequential( # input shape (1, 28, 28)
nn.Conv2d(
in_channels=1, # input height;输入图像的通道数，灰度图像为1，RGB为3
#实际卷积核个数（执行卷积运算的次数）=通道数*filters（卷积核个数）
out_channels=16, # n_filters
kernel_size=5, # filter size；卷积核的长宽
stride=1, # filter movement/step；每次移动卷积核的步长
padding=2, # if want same width and length of this image after Conv2d
#为输入图像外围补充的‘圈数’，每一次卷积核运算的结果为一个特征值
#卷积操作输出的形状计算公式是这样的：
#output_size = (image_size-filter_size+2*padding)/stride + 1
), # output shape (16, 28, 28)
nn.ReLU(), # activation
nn.MaxPool2d(kernel_size=2), # choose max value in 2x2 area, output shape (16, 14, 14)
#2*2的池化，取一个做显著特征
)
#第二层网络
self.conv2 = nn.Sequential( # input shape (16, 14, 14)
nn.Conv2d(16, 32, 5, 1, 2), # output shape (32, 14, 14)
nn.ReLU(), # activation
nn.MaxPool2d(2), # output shape (32, 7, 7)
)
#第三层，全连接层
self.out = nn.Linear(32 * 7 * 7, 10) # fully connected layer, output 10 classes
def forward(self, x):
x = self.conv1(x)
x = self.conv2(x)
x = x.view(x.size(0), -1) # flatten the output of conv2 to (batch_size, 32 * 7 * 7)
#x.size(0)即样本x的个数=60000
#view函数：重新改变张量的维度，参数中的-1就代表这个位置由其他位置的数字来推断（自适应y）
output = self.out(x) #全连接层的输入要为2维张量；规格[60000,y]
return output, x # return x for visualization
cnn = CNN()
#print(cnn) # net architecture
optimizer = torch.optim.Adam(cnn.parameters(), lr=LR) # optimize all cnn parameters
loss_func = nn.CrossEntropyLoss() # the target label is not one-hotted
# following function (plot_with_labels) is for visualization, can be ignored if not interested
from matplotlib import cm
try:
from sklearn.manifold import TSNE #TSNE：随机临近嵌入，主要用于高维数据的降维和可视化
HAS_SK = True #设置一个是否安装该模块的flag,给后面出图做if判断
print(' sklearn installed ')
except:
HAS_SK = False
print('Please install sklearn for layer visualization')
def plot_with_labels(lowDWeights, labels): #lowDWeights[500,2];labels[500]
plt.cla()
X, Y = lowDWeights[:, 0], lowDWeights[:, 1] #2维拆成一维
for x, y, s in zip(X, Y, labels):
c = cm.rainbow(int(255 * s / 9)); plt.text(x, y, s, backgroundcolor=c, fontsize=9)
#设置0-9的颜色；文本颜色及字号
plt.xlim(X.min(), X.max()); plt.ylim(Y.min(), Y.max()); plt.title('Visualize last layer')
plt.show()
plt.pause(0.01)
#设置坐标轴数字显示范围X.min(), X.max() | Y.min(), Y.max();
plt.ion()
# training and testing
for epoch in range(EPOCH):
i = 0
for step, (b_x, b_y) in enumerate(train_loader): # gives batch data, normalize x when iterate train_loader
#每次训练50个数据，样本总数为60000，step=1200
output = cnn(b_x)[0] # cnn output
loss = loss_func(output, b_y) # cross entropy loss
optimizer.zero_grad() # clear gradients for this training step
loss.backward() # backpropagation, compute gradients
optimizer.step() # apply gradients
#随着训练数据的逐步增加，网络逐渐完善，准确率上升。
# 每50步（即每训练完2500个数据）就输入2000个测试数据输出此时网络的准确率
#总共测试60000/2500=24次
if step % 50 == 0:
test_output, last_layer = cnn(test_x) #test_output=output;last_layer=x
#经view()后的last_layer为2维张量；规格[2000,自适应1568]
#print(test_output)
#######
#输出test_output为一个2000*10的张量（取的前两千个测试数据），这10列分别对应的是该测试数据集是0-9的权重
#######
#print(test_output.size())
#print(last_layer.size())
pred_y = torch.max(test_output, 1)[1].data.numpy() #dim=1,输出test_output中每行的最大值的索引
#print(pred_y)
accuracy = float((pred_y == test_y.data.numpy()).astype(int).sum()) / float(test_y.size(0))
#把索引位置等于标签的布尔值true作为整型1求和/标签总数=准确率
i += 1
print("第" + str(i) + "次测试其准确率：")
print('损失: %.4f' % loss.data.numpy(), '，准确率: %.2f' % accuracy)
if HAS_SK:
# Visualization of trained flatten layer (T-SNE)
tsne = TSNE(perplexity=30, n_components=2, init='pca', n_iter=5000) #将输入数据转为2维
plot_only = 500 #取输入数据X的前500个做可视化
low_dim_embs = tsne.fit_transform(last_layer.data.numpy()[:plot_only, :]) # low_dim_embs规格[500,2]
labels = test_y.numpy()[:plot_only] #标签值
plot_with_labels(low_dim_embs, labels)
plt.ioff()
# print 10 predictions from test data
test_output, _ = cnn(test_x[:10])
pred_y = torch.max(test_output, 1)[1].data.numpy()
print(pred_y, 'prediction number')
print(test_y[:10].numpy(), 'real number')
扫一扫
评论
被折叠的 条评论
为什么被折叠?
到【灌水乐园】发言
查看更多评论
添加红包
