bp神经网络添加隐藏层:
import numpy as np
import matplotlib.pyplot as plt
from torch import nn,optim
from torch.autograd import Variable
import torch
from torchvision import datasets,transforms
from torch.utils.data import DataLoader
#导入数据集
train_dataset=datasets.MNIST(root="./",
train=True,
transform=transforms.ToTensor(),
download=False)
test_dataset=datasets.MNIST(root="./",
train=False,
transform=transforms.ToTensor(),
download=False)
#x=train_dataset.test_data
#y=train_dataset.train_labels
#print(x.shape)
#print(y.shape)
#print(train_dataset.train_data.shape)
#plt.imshow(train_dataset.train_data[2],cmap='gray')
#plt.title('%d'%train_dataset.train_labels[2])
#plt.show()
#分批
batch_size=50
input_size=784
hidden_size=500
num_classes=10
train_loader=DataLoader(dataset=train_dataset,
batch_size=batch_size,
shuffle=True)
test_loader=DataLoader(dataset=test_dataset,
batch_size=batch_size,
shuffle=True)
#观察数据维度
for i ,data in enumerate(train_loader):
inputs,labels=data
#print(data.index)
print(inputs.shape) #torch.Size([64,1,28,28])
print(labels.shape) #torch.Size([64])
break
#################################################################################
#实现模型
class SingleNet(nn.Module):
def __init__(self,input_size,hidden_size,num_classes):
super(SingleNet,self).__init__()
self.fc1=nn.Linear(input_size,hidden_size)
self.relu=nn.ReLU()
self.fc2=nn.Linear(hidden_size,num_classes)
def forward(self,x):
out=self.fc1(x)
out=self.relu(out)
out=self.fc2(out)
return out
#################################################################################
model=SingleNet(input_size,hidden_size,num_classes)
loss=nn.CrossEntropyLoss()
optimizer=optim.SGD(model.parameters(),lr=0.01)
#################################################################################
#训练模型
def train():
for i,data in enumerate(train_loader):
inputs,labels=data
inputs=Variable(inputs.view(-1,28*28))
labels=Variable(labels)#把数据变成张量类型
optimizer.zero_grad()
output=model(inputs)
loss1=loss(output,labels)
loss1.backward()#计算损失函数
optimizer.step()
def test():
correct=0
total=0
for i,data in enumerate(test_loader):
inputs,labels=data
inputs=Variable(inputs.view(-1,28*28))
outputs=model(inputs)
_,predicted=torch.max(outputs.data,1)
total+=labels.size(0)
correct+=(predicted==labels).sum()
print("准确率:%d %%"%(torch.floor_divide(100*correct, total)))
##train()
##test()
##train()
##test()
for i in range(10):
train()
test()
无隐藏层:
import numpy as np
import matplotlib.pyplot as plt
from torch import nn,optim
from torch.autograd import Variable
import torch
from torchvision import datasets,transforms
from torch.utils.data import DataLoader
#导入数据集
train_dataset=datasets.MNIST(root="./",
train=True,
transform=transforms.ToTensor(),
download=False)
test_dataset=datasets.MNIST(root="./",
train=False,
transform=transforms.ToTensor(),
download=False)
#数据太多,不要分批训练,定义批次生成器,和每个批次大小
batch_size=50
train_loader=DataLoader(dataset=train_dataset,
batch_size=batch_size,
shuffle=True)
test_loader=DataLoader(dataset=test_dataset,
batch_size=batch_size,
shuffle=True)
for i ,data in enumerate(train_loader):
inputs,labels=data
print(inputs.shape) #torch.Size([64,1,28,28])
print(labels.shape) #torch.Size([64])
break
print(labels)#0-9的数据,总共一批64个
print(len(train_loader),len(test_loader))#总共会循环938次60000张图片,分成938个批次,每个批次64个值
########################################################
#########定义网络########################
class Net(nn.Module):
def __init__(self):
super(Net,self).__init__()#父类初始化,固定格式
self.fcl=nn.Linear(784,10)#定义一个简单神经网络输入层直接到输出层fcl
self.softmax=nn.Softmax(dim=1) #定义激活幻术softmax.dim是维度从0开始,等于1就是矩阵的列
def forward(self,x):#前向传播函数
#x是([64,1,28,28])需要变成([64,784])
x=x.view(x.size()[0],-1)#0是64,-1是自动匹配1*28*28=784
x=self.fcl(x) #传给网络再激活返回
x=self.softmax(x)
return x
########################################################
####定义模型
model=Net()
####定义代价函数MSEloss均方损失函数,分类问题用交叉熵比较好
#mse_loss=nn.MSELoss()
mse_loss=nn.CrossEntropyLoss()
####定义优化器,最简单的梯度下降法SGD,需要传入模型参数
optimizer=optim.SGD(model.parameters(),0.5)
######################################################
##定义模型训练函数
##
def train():
for i,data in enumerate(train_loader):#循环938次
#获得一个批次数据
inputs,labels=data
out=model(inputs)#输出是(64,10)
#交叉熵代价函数out是(batch.C)labels(batch)这个损失函数以可两维对一维,因为独热编码已经被封装在函数内部了
loss=mse_loss(out,labels)
#梯度清零
optimizer.zero_grad()
#计算梯度
loss.backward()
#跟新权值
optimizer.step()
def test():
correct=0
for i,data in enumerate(test_loader):
inputs,labels=data
out=model(inputs)
#计算out的1维度最大值所在的位置,softmax()所在计算出的最大概率所在的位置
_,predicted=torch.max(out,1)
#预测值和标签值相同的数目累加的
correct+=(predicted==labels).sum()
print("准确率:{0}".format(correct.item()/len(test_dataset)))
for i in range(10):
print('i:',i)
train()
test()
##train()
##test()
##train()
##test()
![在这里插入图片描述](https://img-blog.csdnimg.cn/20201109151629870.png?x-oss-process=image/watermark,type_ZmFuZ3poZW5naGVpdGk,shadow_10,text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L1N3ZWV0Qm9uaw==,size_16,color_FFFFFF,t_70#pic_center)