动手学深度学习（一、深度学习基础--线性回归）

动手学深度学习（一、深度学习基础--线性回归）

动手学深度学习（Pytorch版本）

生成数据集

读取数据

初始化模型参数

定义模型

定义损失函数

定义优化算法

训练模型

一、线性回归从零开始

import torch
from IPython import display
from matplotlib import pyplot as plt
import numpy as np
import random

num_inputs = 2
num_examples = 1000
true_w = [2, -3.4]
true_b = 4.2
features = torch.randn(num_examples, num_inputs, dtype=torch.float32)
labels = true_w[0] * features[:, 0] + true_w[1] * features[:, 1] + true_b
labels += torch.tensor(np.random.normal(0, 0.01, size=labels.size()), dtype=torch.float32)
print(features[0], labels[0])

#===========================matplotlib画图技术
def use_svg_display():
    display.set_matplotlib_formats('svg')
    
def set_figsize(figsize = (3.5, 2.5)):
    use_svg_display()
    plt.rcParams['figure.figsize'] = figsize
    
set_figsize()
plt.scatter(features[:, 1].numpy(), labels.numpy(), 1);
#==================================================

def data_iter(batch_size, features, labels):
    num_examples = len(features)
    indices = list(range(num_examples))
    
    random.shuffle(indices)
    for i in range(0, num_examples, batch_size):
        j = torch.LongTensor(indices[i: min(i + batch_size, num_examples)]) # 最后一次可能不足一个batch
        yield  features.index_select(0, j), labels.index_select(0, j)
        
batch_size = 10
for X, y in data_iter(batch_size, features, labels):
    print(X, y)
    break
    
w = torch.tensor(np.random.normal(0, 0.01, (num_inputs, 1)), dtype=torch.float32)
b = torch.zeros(1, dtype=torch.float32)
print(w)
print(b)
w.requires_grad_(requires_grad=True)
b.requires_grad_(requires_grad=True) 

#定义模型
def linreg(X, w, b):
    return torch.mm(X, w) + b

#定义损失函数
def squared_loss(y_hat, y):
    return (y_hat - y.view(y_hat.size())) ** 2 / 2

#定义优化算法
def sgd(params, lr, batch_size):
    for param in params:
        param.data -= lr * param.grad / batch_size
        
lr = 0.03
num_epochs = 3
net = linreg
loss = squared_loss

for epoch in range(num_epochs):
    for X, y in data_iter(batch_size, features, labels):
        l = loss(net(X, w, b), y).sum()
        l.backward()
        sgd([w, b], lr, batch_size)
        
        w.grad.data.zero_()
        b.grad.data.zero_()
    
    train_l = loss(net(features, w, b), labels)
    print('epoch %d, loss %f' %(epoch + 1, train_l.mean().item()))
    

print(true_w, '\n', w)
print(true_b, '\n', b)

tensor([ 1.7489, -1.3365]) tensor(12.2518)
tensor([[-0.7026, -0.9066],
        [-0.5462,  0.7684],
        [ 0.1347, -1.1432],
        [-0.9276,  0.2061],
        [-0.2879,  0.8351],
        [-0.1709, -0.2726],
        [ 0.3896,  0.6176],
        [ 0.9559, -0.9961],
        [ 0.3178, -0.7674],
        [ 0.1679, -0.2784]]) tensor([5.8999, 0.4829, 8.3593, 1.6556, 0.7915, 4.7864, 2.8755, 9.5106, 7.4385,
        5.4885])
tensor([[-0.0004],
        [ 0.0188]])
tensor([0.])
epoch 1, loss 0.039256
epoch 2, loss 0.000147
epoch 3, loss 0.000051
[2, -3.4] 
 tensor([[ 1.9998],
        [-3.3993]], requires_grad=True)
4.2 
 tensor([4.1995], requires_grad=True)

二、线性回归简洁实现

import torch
from matplotlib import pyplot as plt
import numpy as np
import random

#1.生成数据集
num_inputs = 2
num_examples = 1000
true_w = [2, -3.4]
true_b = 4.2
#正态分布
features = torch.tensor(np.random.normal(0, 1, (num_examples, num_inputs)), dtype=torch.float)
labels = true_w[0] * features[:, 0] + true_w[1] * features[:, 1] + true_b
labels += torch.tensor(np.random.normal(0, 0.01, size=labels.size()), dtype=torch.float)

print(features[0], labels[0])

#2.读取数据
import torch.utils.data as Data
batch_size = 10
#将训练数据的特征和标签组合
dataset = Data.TensorDataset(features, labels)
#随机读取小批量
data_iter = Data.DataLoader(dataset, batch_size, shuffle = True)
for X, y in data_iter:
    print(X)
    print(y)
    break

#3.定义模型
import torch.nn as nn
class LinearNet(nn.Module):
    def __init__(self, n_feature):
        super(LinearNet, self).__init__()
        self.linear = nn.Linear(n_feature, 1)
    #forward定义前向传播
    def forward(self, x):
        y = self.linear(x)
        return y

net = LinearNet(num_inputs)
print(net)

for param in net.parameters():
    print(param)
    
#4.初始化模型参数， 初始化模型中的权重和偏差
from torch.nn import init
#初始化模型的权重和偏差
init.normal_(net.linear.weight, mean = 0, std = 0.01)
init.constant_(net.linear.bias, val = 0)
#net[0].weight应改为net.linear.weight， 因为net[0]这样根据下标访问子模块的写法只有当net是个ModuleList或者Sequential实例时才可以
for param in net.parameters():
    print(param)
    
#5.定义损失函数
loss = nn.MSELoss()

#6.定义优化算法（新建优化器会造成损失函数的收敛出现震荡的情况）
import torch.optim as optim
optimizer = optim.SGD(net.parameters(), lr = 0.03)
print(optimizer)
#可以为不同子网络设置不同的学习率， 这在finetune时经常用到
#finetune微调

#调整学习率
print("*" * 20)
print(len(optimizer.param_groups))
for param_group in optimizer.param_groups:
    param_group['lr'] *= 1

#7.训练模型
#我们通过调用optim实例的step函数来迭代模型参数。按照小批量随机梯度下降的定义，我们在step函数中指明批量大小，从而对批量中样本梯度求平均。

num_epochs = 3
for epoch in range(1, num_epochs + 1):
    for X, y in data_iter:
        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 %d, loss: %f'%(epoch, l.item()))
    
#我们从net获得需要的层，访问其权重（weight）和偏差（bias）
dense = net.linear
print(true_w, dense.weight)
print(true_b, dense.bias)

#使用Pytorch可以更简洁地实现模型
#torch.utils.data提供了有关数据处理的工具
# torch.nn模块定义了大量神经网络的层
# torch.nn.init模块定义了各种初始化方法
# torch.optim模块提供了很多常用的优化算法

tensor([0.8380, 1.8332]) tensor(-0.3578)
tensor([[-0.5930, -1.0323],
        [ 0.4803, -0.6949],
        [-1.8518, -0.3402],
        [ 2.1459, -0.9358],
        [ 0.0538,  0.1641],
        [-0.2422,  2.7112],
        [-1.4170, -0.2609],
        [ 1.1335,  1.3915],
        [-0.9495,  0.0463],
        [ 0.7052, -0.1217]])
tensor([ 6.5231,  7.5113,  1.6531, 11.6631,  3.7448, -5.5022,  2.2547,  1.7412,
         2.1505,  6.0146])
LinearNet(
  (linear): Linear(in_features=2, out_features=1, bias=True)
)
Parameter containing:
tensor([[-0.0726,  0.3277]], requires_grad=True)
Parameter containing:
tensor([-0.6877], requires_grad=True)
Parameter containing:
tensor([[-0.0027, -0.0005]], requires_grad=True)
Parameter containing:
tensor([0.], requires_grad=True)
SGD (
Parameter Group 0
    dampening: 0
    lr: 0.03
    momentum: 0
    nesterov: False
    weight_decay: 0
)
********************
1
epoch 1, loss: 0.000124
epoch 2, loss: 0.000123
epoch 3, loss: 0.000023
[2, -3.4] Parameter containing:
tensor([[ 1.9998, -3.3999]], requires_grad=True)
4.2 Parameter containing:
tensor([4.2010], requires_grad=True)