池化层有双重目的：降低卷积层对位置的敏感性，同时降低对空间降采样表示的敏感性

import torch
from torch import nn
from d2l import torch as d2l


# 实现池化层的正向传播
def pool2d(X, pool_size, mode='max'):
    p_h, p_w = pool_size  # 池化层窗口大小
    Y = torch.zeros((X.shape[0] - p_h + 1, X.shape[1] - p_w + 1))
    for i in range(Y.shape[0]):
        for j in range(Y.shape[1]):
            if mode == 'max':
                Y[i, j] = X[i: i + p_h, j: j + p_w].max()  # 最大汇聚(池化)
            elif mode == 'avg':
                Y[i, j] = X[i: i + p_h, j: j + p_w].mean()  # 平均汇聚(池化)
    return Y


# 验证
X = torch.tensor([[0.0, 1.0, 2.0], [3.0, 4.0, 5.0], [6.0, 7.0, 8.0]])
print(pool2d(X, (2, 2)))
print(pool2d(X, (2, 2), mode='avg'))

# 填充和步幅
X = torch.arange(16, dtype=torch.float32).reshape((1, 1, 4, 4))
print(X)

# 默认步幅与汇聚窗口的大小相同
pool2d = nn.MaxPool2d(3)
print(pool2d(X))

# 填充和步幅可以手动设定
pool2d = nn.MaxPool2d(3, padding=1, stride=2)
print(pool2d(X))

# 若有多个通道，池化层在每个输入通道上单独运算
X = torch.cat((X, X + 1), 1)
print(X)
pool2d = nn.MaxPool2d(3, padding=1, stride=2)
print(pool2d(X))