- 导入包:
- 数据集的生成
- 数据的读取
- 定义模型
- 初始化模型参数
- MSE损失函数
- 定义优化算法
- 模型的优化
线性回归详细实现, 请点击此处 导入包:
import torch
import numpy as np
import torch.utils.data as Data # 数据读取
import torch.nn as nn
from torch.nn import init # 初始化
import torch.optim as optim # 优化器
数据集的生成
num_inputs = 2 # 2个维度
num_examples = 1000 # 1000调数据
# 标准的参数
true_w = [2, -3.4]
true_b = 4.2
features = torch.tensor(np.random.normal(0, 1, (num_examples,num_inputs)), dtype=torch.float)
label = true_w[0] * features[:, 0] + true_w[1] * features[:, 1] +true_b
label += torch.tensor(np.random.normal(0, 0.01,size=label.size()), dtype=torch.float)
数据的读取
batch_size = 30
dataset = Data.TensorDataset(features,label)
data_it = Data.DataLoader(dataset,batch_size,shuffle =True)
定义模型
定义模型有多种方法:
方法一:继承nn.Module
class LinearNet(nn.Module):
def __init__(self,n_feature):
super(LinearNet, self).__init__()
self.linear = nn.Linear(n_feature,1)
def forward(self, x):
y = self.linear(x)
return y
net = LinearNet(num_inputs)
for param in net.parameters():
print(param)
方法二:nn.Sequential
net = nn.Sequential(
nn.Linear(num_inputs, 1)
)
方法三:nn.Sequential()+add_module
net = nn.Sequential()
net.add_module('linear', nn.Linear(num_inputs, 1))
方法四:导入OrderedDict
from collections import OrderedDict
net = nn.Sequential(OrderedDict([
('linear', nn.Linear(num_inputs, 1))]))
初始化模型参数
init.normal_(net[0].weight, mean=0, std=0.01)
init.constant_(net[0].bias,val=2)
MSE损失函数
loss = nn.MSELoss()
定义优化算法
optimizer = optim.SGD(net.parameters(), lr=0.03)
模型的优化
num_epochs = 3
for epoch in range(1, num_epochs + 1):
for X, y in data_it:
output = net(X)
l = loss(output, y.view(-1, 1))
optimizer.zero_grad() # 梯度清零,等价于net.zero_grad()
l.backward()
optimizer.step()
print(epoch, l.item())
最后输出epoch和loss:
1 0.3572668433189392
2 0.005662666633725166
3 0.00011592111695790663
喜欢文章可以点赞收藏,欢迎关注,如有错误请指正!
欢迎分享,转载请注明来源:内存溢出
评论列表(0条)