一、感知机基本原理
二、代码实现
1.导入第三方库
所需库均为常用机器学习中的常用库,其中使用tensorboard进行绘图,具体操作可参考博文https://blog.csdn.net/yhl1001/article/details/131296005
如不习惯读者可自行更换为matplotlib
import torch
import torchvision
from torch import nn
from torch.utils import data
from torchvision import transforms
from torch.utils.tensorboard import SummaryWriter
2.导入数据集
本文感知机实现使用的是FashionMNIST数据集,运行下方代码可直接进行下载。如果时间过长也可通过下方网址自行下载后导入
训练数据集:http://fashion-mnist.s3-website.eu-central-1.amazonaws.com/train-images-idx3-ubyte.gz
测试数据集:
http://fashion-mnist.s3-website.eu-central-1.amazonaws.com/test-images-idx3-ubyte.gz
这个数据集中包含10类
train_data = torchvision.datasets.FashionMNIST(root='./data', transform=transforms.ToTensor(), download=True, train=True)
test_data = torchvision.datasets.FashionMNIST(root='./data', transform=transforms.ToTensor(), download=True, train=False)
3.使用dataloader
利用DataLoader生成一个小批量提取数据迭代器。(如果运行不成功,将num_worker改为=0)
train_loader = data.DataLoader(train_data, 64, shuffle=True, num_workers=4)
test_loader = data.DataLoader(test_data, 64, shuffle=True, num_workers=4)
4.构建神经网络
构建一个神经网络,由两层线性网络构成,第一层使用relu激活函数,第二层则是softmax。
#定义神经网络
class MLP(nn.Module):
def __init__(self):
super().__init__()
self.sequential = nn.Sequential(
nn.Flatten(),
nn.Linear(784, 256),
nn.ReLU(),
nn.Linear(256, 10),
nn.Softmax()
)
def forward(self, x):
x = self.sequential(x)
return x
mlp = MLP()
5.定义损失函数,和测准率,优化器
损失函数使用交叉熵,测准率为 正确分类个数/总数
优化器使用SGD,大家可尝试加入权重衰退
# 定义损失函数
loss_fn = nn.CrossEntropyLoss()
def accuracy(y, y_hat):
ac = y == y_hat
ac = ac.sum()/len(y)
return ac.item()
a = torch.tensor([0, 1])
b = torch.tensor([1, 1])
accuracy(a, b)
trainer = torch.optim.SGD(yhl.parameters(), lr=0.03)
6.进行训练,并绘制图像
这里我们迭代5次,精度大概在80%左右
epochs = 5
writer = SummaryWriter("logs")
for epoch in range(epochs):
print(f"正在进行第{epoch+1}次迭代")
train_loss = 0
test_loss = 0
train_ac = 0
test_ac = 0
m, n = 0, 0
for x, y in train_loader:
y_hat = mlp(x)
loss = loss_fn(y_hat, y)
train_loss += loss
_, y_hat= y_hat.max(1)
ac = accuracy(y, y_hat)
n += 1
train_ac += ac
trainer.zero_grad()
loss.backward()
trainer.step()
writer.add_scalar("train_accuaracy",train_ac/n, epoch+1)
print(train_loss.item()/n, train_ac/n)
for x, y in test_loader:
y_hat = yhl(x)
loss = loss_fn(y_hat, y)
test_loss += loss
_, y_hat= y_hat.max(1)
ac = accuracy(y, y_hat)
test_ac += ac
m += 1
writer.add_scalar("test_accuaracy",test_ac/n, epoch+1)
print(test_ac/m)
writer.close()
有什么问题欢迎大家私信,或者在评论区留言。