python卷积网络进行识别，手写字符识别

gaohaikuo1

batch size 含义https://blog.csdn.net/qq_42722197/article/details/129828503

gaohaikuo1

使用Conv2d进行简单二维卷积运算
import torch
import torch.nn as nn
# With square kernels and equal stride
m= nn.Conv2d(1, 1, 1, stride=1)
input = torch.randn(1,1,2,2)
input = torch.tensor([[[[ 1., 2.],[ 3., 4.]]]])
m.weight = torch.nn.Parameter(torch.ones(1,1,1,1))
m.bias = torch.nn.Parameter(torch.zeros(1))
output = m(input)
print(input)
print(m.weight)
print(m.bias)
print(output)

gaohaikuo1

import torch
import torch.nn as nn
# With square kernels and equal stride
m= nn.Conv2d(1, 1, 1, stride=1)
input = torch.randn(1,1,2,2)
input = torch.tensor([[[[ 1., 2.],[ 3., 4.]]]])
m.weight = torch.nn.Parameter(torch.ones(1,1,1,1))
output = m(input)
print(input)
print(m.weight)
print(output)

gaohaikuo1

Pytorch 中 torch.nn.Conv2d和torch.nn.functional.conv2d的区别
https://blog.csdn.net/Z2572862506/article/details/128761206
当我们想要自定义卷积核时，是不能使用 torch.nn.ConvNd 的，因为它里面的权重都是需要学习的参数，没有办法自行定义。但是，我们可以使用 torch.nn.functional.conv2d()

gaohaikuo1

接上，如何自定义
import torch
import torch.nn as nn

conv = nn.Conv2d(in_channels=2, out_channels=1, kernel_size=3)
# 此时weight以及bias已由nn.Conv2d初始化
conv.weight, conv.bias
# 手动设定
# conv.weight.data 以及 conv.bias.data属性为torch.tensor
# 因此只要获取conv.weight.data以及conv.bias.data属性，后续调用torch.tensor的不同方法即可进行修改
# 例如：全部修改为0
conv.weight.data.zero_(), conv.bias.data.zero_()
conv.weight, conv.bias

https://www.jb51.net/article/274016.htm