詳解部分：
1.Dropout是為了防止過擬合而設置的
2.Dropout顧名思義有丟掉的意思
3.nn.Dropout(p = 0.3) # 表示每個神經(jīng)元有0.3的可能性不被激活
4.Dropout只能用在訓練部分而不能用在測試部分
5.Dropout一般用在全連接神經(jīng)網(wǎng)絡映射層之后，如代碼的nn.Linear(20, 30)之后

在這里插入圖片描述

代碼部分：

class Dropout(nn.Module):
	def __init__(self):
		super(Dropout, self).__init__()
		self.linear = nn.Linear(20, 40)
		self.dropout = nn.Dropout(p = 0.3) # p=0.3表示下圖（a）中的神經(jīng)元有p = 0.3的概率不被激活

	def forward(self, inputs):
		out = self.linear(inputs)
		out = self.dropout(out)
		return out

net = Dropout()
# Dropout只能用在train而不能用在test

2.nn.Dropout用法二

以代碼為例

import torch
import torch.nn as nn
a = torch.randn(4, 4)
print(a)
"""
tensor([[ 1.2615, -0.6423, -0.4142,  1.2982],
        [ 0.2615,  1.3260, -1.1333, -1.6835],
        [ 0.0370, -1.0904,  0.5964, -0.1530],
        [ 1.1799, -0.3718,  1.7287, -1.5651]])
"""
dropout = nn.Dropout()
b = dropout(a)
print(b)
"""
tensor([[ 2.5230, -0.0000, -0.0000,  2.5964],
        [ 0.0000,  0.0000, -0.0000, -0.0000],
        [ 0.0000, -0.0000,  1.1928, -0.3060],
        [ 0.0000, -0.7436,  0.0000, -3.1303]])
"""

由以上代碼可知Dropout還可以將部分tensor中的值置為0

補充：torch.nn.dropout和torch.nn.dropout2d的區(qū)別

import torch
import torch.nn as nn
import torch.autograd as autograd

m = nn.Dropout(p=0.5)
n = nn.Dropout2d(p=0.5)
input = autograd.Variable(torch.randn(1, 2, 6, 3)) ## 對dim=1維進行隨機置為0

print(m(input))
print('****************************************************')
print(n(input))

在這里插入圖片描述

下面的都是錯誤解釋和錯誤示范，沒有刪除的原因是留下來進行對比，希望不要犯這類錯誤

# -*- coding: utf-8 -*-
import torch
import torch.nn as nn
import torch.autograd as autograd

m = nn.Dropout(p=0.5)
n = nn.Dropout2d(p=0.5)
input = autograd.Variable(torch.randn(2, 6, 3)) ## 對dim=1維進行隨機置為0

print(m(input))
print('****************************************************')
print(n(input))

結(jié)果是：