Python PyTorch 中的 torch.nn.Dropout() 方法

2025 年 5 月 27 日 | 4 分钟阅读

在训练过程中，torch.nn.Dropout() 会以给定概率将输入的某些部分设为零。通过正则化采用此技术可以防止过拟合。但是，需要注意的是，此方法仅适用于非复数值输入。

语法

它具有以下语法：

参数

p（浮点型，可选）：将元素置零的概率。默认值为 0.5。
inplace（布尔型，可选）：如果设置为 TRUE，则操作在原地完成。默认值为 False。

示例 1

在此示例中，我们使用 torch.nn.Dropout()，输入概率为 0.35，这意味着输入张量中的每个元素将有 35% 的概率被置零。

代码

# Import required library
import torch

# Define a float tensor
tens = torch.tensor([-0.7345, 0.4347, -0.1237, 1.3379, 0.2343], dtype=torch.float32)

# Print the tensor
print("Original tensor:", tens)

# probability p=0.35
drop = torch.nn.Dropout(0.35)

# Apply dropout (in training mode)
drop.train()  # This is default, but added for clarity
output_tens = drop(tens)

# output
print("Output tensor:", output_tens)   

输出

Original tensor: tensor([-0.7345,  0.4347, -0.1237,  1.3379,  0.2343]) 
Output tensor: tensor([-1.1308,  0.0000, -0.1903,  2.0583,  0.3605])

示例 2

在这种情况下，我们使用 torch.nn.Dropout() 方法，概率为 (0.85)，且 inplace = True。输入张量的每个元素有 85% 的概率被置零，并且是原地操作。

代码

import torch

# Define a float tensor
tens = torch.tensor([[-0.1345, -0.7437, 1.2377],
                     [0.9337, 1.6473, 0.4346],
                     [-0.6345, 0.9344, -0.2456]], dtype=torch.float32)

# Print the original tensor
print("\nOriginal tensor:\n", tens)

# Use torch.nn.Dropout() with p=0.85 and inplace=True
drop = torch.nn.Dropout(p=0.85, inplace=True)

# Apply dropout (in training mode by default)
drop(tens)

# Display the modified tensor (in-place)
print("\nOutput tensor after applying Dropout:\n", tens)   

输出

Original tensor:
 tensor([[-0.1345, -0.7437,  1.2377],
        [ 0.9337,  1.6473,  0.4346],
        [-0.6345,  0.9344, -0.2456]])

Output tensor after applying dropout:
 tensor([[ 0.0000,  0.0000,  0.0000],
        [ 0.0000, 10.9820,  0.0000],
        [ 0.0000,  0.0000, -1.6372]])

示例 3

在下一个 Python 程序中，我们将 p = 0.5；因此，每个元素将有 50% 的概率被置零。

代码

Code:
import torch

# Define a tensor with 4 random elements
tensor = torch.randn(4)

print("Input tensor:", tensor)

# Dropout layer with p=0.5
dropout = torch.nn.Dropout(0.5)

# Apply dropout
output = dropout(tensor)

print("Output Tensor:", output)   

输出

Input tensor: tensor([ 0.3667, -0.9647, 0.8537, 1.3171]) 
Output Tensor: tensor([ 0.7334, -1.9294, 0.0000, 0.0000])

示例 4

在下面的 Python 程序中，我们定义 p = 0.25；也就是说，输入张量中的任何元素有 25% 的概率被置零。

代码

import torch

# Define a tensor with random values of size 4x3
tensor = torch.randn(4, 3)

# Print the input tensor
print("Input tensor:", tensor)

# Define a dropout layer with probability p=0.25
dropout = torch.nn.Dropout(0.25)

# Apply the dropout layer to the tensor
output = dropout(tensor)

# Print the output tensor after applying dropout
print("Output Tensor:", output)   

输出

Input tensor: tensor([[-0.0999,  0.3457,  0.4531],
                      [ 1.2090, -0.7589,  1.0932],
                      [-0.1260,  0.7774,  1.0710],
                      [ 1.4131, -1.7814, -0.5732]])

Output Tensor: tensor([[-0.1331,  0.4609,  0.6042],
                       [ 1.6120, -1.0119,  1.4576],
                       [-0.0000,  1.0365,  1.4280],
                       [ 1.8841, -0.0000, -0.7642]])

示例 5

import numpy as np
import pandas as pd
import torch
import torch.nn as nn
import torch.optim as optim
from sklearn.preprocessing import LabelEncoder
from sklearn.model_selection import StratifiedKFold

# Read data
data = pd.read_csv("sonar.csv", header=None)
X = data.iloc[:, 0:60]
y = data.iloc[:, 60]

# string to integer
encoder = LabelEncoder()
encoder.fit(y)
y = encoder.transform(y)

# Convert to 2D PyTorch tensors
X = torch.tensor(X.values, dtype=torch.float32)
y = torch.tensor(y, dtype=torch.float32).reshape(-1, 1)

# Define PyTorch model
class SonarModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.layer1 = nn.Linear(60, 60)
        self.act1 = nn.ReLU()
        self.layer2 = nn.Linear(60, 30)
        self.act2 = nn.ReLU()
        self.output = nn.Linear(30, 1)
        self.sigmoid = nn.Sigmoid()

    def forward(self, x):
        x = self.act1(self.layer1(x))
        x = self.act2(self.layer2(x))
        x = self.sigmoid(self.output(x))
        return x

# Helper function
def model_train(model, X_train, y_train, X_val, y_val,
                n_epochs=300, batch_size=16):
    loss_fn = nn.BCELoss()
    optimizer = optim.SGD(model.parameters(), lr=0.01, momentum=0.8)
    batch_start = torch.arange(0, len(X_train), batch_size)

    model.train()
    for epoch in range(n_epochs):
        for start in batch_start:
            X_batch = X_train[start:start+batch_size]
            y_batch = y_train[start:start+batch_size]
            y_pred = model(X_batch)
            loss = loss_fn(y_pred, y_batch)
            optimizer.zero_grad()
            loss.backward()
            optimizer.step()

    # Evaluate accuracy after training
    model.eval()
    y_pred = model(X_val)
    acc = (y_pred.round() == y_val).float().mean()
    acc = float(acc)
    return acc

# run 10-fold cross validation
kfold = StratifiedKFold(n_splits=10, shuffle=True)
accuracies = []
for train, test in kfold.split(X, y):
    # create model, train, and get accuracy
    model = SonarModel()
    acc = model_train(model, X[train], y[train], X[test], y[test])
    print("Accuracy: %.2f" % acc)
    accuracies.append(acc)

# evaluating
mean = np.mean(accuracies)
std = np.std(accuracies)
print("Baseline: %.2f%% (+/- %.2f%%)" % (mean*100, std*100))   

输出

Accuracy: 0.79
Accuracy: 0.83
Accuracy: 0.76
Accuracy: 0.81
Accuracy: 0.78
Accuracy: 0.80
Accuracy: 0.75
Accuracy: 0.82
Accuracy: 0.77
Accuracy: 0.79
Baseline: 79.10% (+/- 2.72%)

下一主题主成分分析

Python PyTorch 中的 torch.nn.Dropout() 方法

语法

示例 1

示例 2

示例 3

示例 4

示例 5

联系信息

关注我们

教程

面试题

在线编译器

Python

Java

.Net Framework

AI, ML and Data Science

Cloud Technology

B.Tech and MCA

Web Technology

PHP

Software Testing

Technical Interview

Java Interview

Python

Web Interview

Database Interview

B.Tech / MCA

Important Interview

Software Testing Interview

Company Interviews

Online Compilers

Multiple Choice Questions

机器学习

监督式学习

分类

杂项

相关教程

面试题

Python PyTorch 中的 torch.nn.Dropout() 方法

语法

示例 1

示例 2

示例 3

示例 4

示例 5

相关帖子

机器学习中的矩阵分解

机器学习在信号处理中的应用

最大似然估计简介

机器学习在国防/军事领域的应用

决策树分类算法

用于推荐系统的矩阵分解

机器学习中的因果关系简介

机器人是否是自学的？

什么是人机协作机器学习

如何在 Python 中计算变量之间的相关性

订阅 Tpoint Tech

联系信息

关注我们

教程

面试题

在线编译器