disout

最新推荐文章于 2024-08-25 07:43:25 发布

AI算法网奇

最新推荐文章于 2024-08-25 07:43:25 发布

阅读量4.5k

点赞数

分类专栏：深度学习宝典

本文链接：https://blog.csdn.net/jacke121/article/details/106219156

版权

深度学习宝典专栏收录该内容

988 篇文章 714 订阅 ¥29.90 ¥99.00

订阅专栏

超级会员免费看

Disout是一种新颖的深度学习正则化技术，通过在特征图上引入可控的噪声来分散注意力，提高模型的泛化能力。该技术借鉴了DropBlock的思想，但提出了一种更平滑的噪声分布方式，减少了训练过程中的方差。Disout在多个计算机视觉任务上表现出优于dropout和其他正则化方法的性能。

摘要由CSDN通过智能技术生成

参考：

https://github.com/huawei-noah/Disout/blob/1c8591f36240b8591691a972c19c7f96134b52a5/disout.py

#Copyright (C) 2020. Huawei Technologies Co., Ltd. All rights reserved.

#This program is free software; you can redistribute it and/or modify it under the terms of the BSD 3-Clause License.

#This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the BSD 3-Clause License for more details.


import torch.nn as nn
import torch
import torch.nn.functional as F
import numpy as np

class Disout(nn.Module):