一种神经网络模型

import torch
import torch.nn as nn
import torch.nn.functional as F

#定义一个网络
class Net(nn.Module):

    def __init__(self):
        super(Net, self).__init__()
        # 1 input image channel, 6 output channels, 5x5 square convolution
        # kernel
        self.conv1 = nn.Conv2d(1, 6, 5)
        # print("----卷积核----")
        # print(self.conv1)
        self.conv2 = nn.Conv2d(6, 16, 5)
        # an affine operation: y = Wx + b
        self.fc1 = nn.Linear(16 * 5 * 5, 120)
        self.fc2 = nn.Linear(120, 84)
        self.fc3 = nn.Linear(84, 10)

    def forward(self, x):
        # Max pooling over a (2, 2) window
        x = F.max_pool2d(F.relu(self.conv1(x)), (2, 2))
        # If the size is a square you can only specify a single number
        x = F.max_pool2d(F.relu(self.conv2(x)), 2)
        x = x.view(-1, self.num_flat_features(x))
        x = F.relu(self.fc1(x))
        x = F.relu(self.fc2(x))
        x = self.fc3(x)
        return x

    def num_flat_features(self, x):
        size = x.size()[1:]  # all dimensions except the batch dimension
        num_features = 1
        for s in size:
            num_features *= s
        return num_features

if __name__ == "__main__":
    net = Net()
    print(net)

学习torch框架中的卷积神经网络

一，nn.Conv2d

二维卷积可以处理二维数据
1，nn.Conv2d(self, in_channels, out_channels, kernel_size, stride=1, padding=0, dilation=1, groups=1, bias=True))
参数：
self:

in_channels:输入数据的通道数，例RGB图片通道数为3；

out_channel:输出数据的通道数，这个根据模型调整；

kernel_size:卷积核大小，可以是int，或tuple；kennel_size=5,意味着卷积大小(5,5).

stride: 步长，默认为1，与kennel_size类似，stride=2,意味着步长上下左右扫描皆为2， tride=（2,3），左右扫描步长为2，上下为3；

padding：　零填充

dilation:

groups:
bias:

2例子 输入数据X[10,16,30,32],其分别代表：10组数据，通道数为16，高度为30，宽为32

import torch
import torch.nn as nn
import torch.nn.functional as F

x = torch.randn(10, 16, 30, 32, 34)
# batch, channel , height , width
print(x.shape)
class Net_1D(nn.Module):
    def __init__(self):
        super(Net_1D, self).__init__()
        self.layers = nn.Sequential(
            nn.Conv1d(in_channels=16, out_channels=16, kernel_size=(3, 2, 2), stride=(2, 2, 1), padding=[2,2,2]),
            nn.ReLU()
        )
    def forward(self, x):
        output = self.layers(x)
        log_probs = F.log_softmax(output, dim=1)
        return  log_probs

n = Net_1D()  # in_channel,out_channel,kennel,
print(n)
y = n(x)
print(y.shape)

结果：

torch.Size([10, 16, 30, 32, 34])
Net_1D(
  (layers): Sequential(
    (0): Conv1d(16, 16, kernel_size=(3, 2, 2), stride=(2, 2, 1), padding=[2, 2, 2])
    (1): ReLU()
  )
)
torch.Size([10, 16, 16, 18, 37])

3.卷积计算
d = (d - kennel_size + 2 * padding) / stride + 1
x = ([10,16,30,32,34]),其中第一维度：30，第一维度,第二维度：32,第三维度：34，对于卷积核长分别是；对于步长分别是第一维度：2,第二维度：,2,第三维度：1；对于padding分别是：第一维度：2,第二维度：,2,第三维度：2；
d1 = (30 - 3 + 22)/ 2 +1 = 31/2 +1 = 15+1 =16
d2 = (32 - 2 + 22)/ 2 +1 = 34/2 +1 = 17+1 =18
d3 = (34 - 2 + 2*2)/ 1 +1 = 36/1 +1 = 36+1 =37
batch = 10, out_channel = 16

故：y = [10, 16, 16, 18, 37]

二，nn.Linear()

1,PyTorch的nn.Linear（）是用于设置网络中的全连接层的，需要注意的是全连接层的输入与输出都是二维张量，一般形状为[batch_size, size]，不同于卷积层要求输入输出是四维张量。其用法与形参说明如下：

torch.nn.Linear(in_features,out_features,bias=True)
in_features:指的是输入的二维张量的大小，即输入的[batch_size, size]中的size。

out_features:指的是输出的二维张量的大小，即输出的二维张量的形状为[batch_size，output_size]，当然，它也代表了该全连接层的神经元个数。从输入输出的张量的shape角度来理解，相当于一个输入为[batch_size, in_features]的张量变换成了[batch_size, out_features]的输出张量。

2,用法示例：

import torch as t
from torch import nn

# in_features由输入张量的形状决定，out_features则决定了输出张量的形状 
connected_layer = nn.Linear(in_features = 64*64*3, out_features = 1)

# 假定输入的图像形状为[64,64,3]
input = t.randn(1,64,64,3)

# 将四维张量转换为二维张量之后，才能作为全连接层的输入
input = input.view(1,64*64*3)
print(input.shape)
output = connected_layer(input) # 调用全连接层
print(output.shape)

这段代码运行结果为：

input shape is %s torch.Size([1, 12288])
output shape is %s torch.Size([1, 1])