A Unified Contrastive Loss for Self-Training

Aurelien Gauffre, Julien Horvat, Massih-Reza Amini
{"title":"A Unified Contrastive Loss for Self-Training","authors":"Aurelien Gauffre, Julien Horvat, Massih-Reza Amini","doi":"arxiv-2409.07292","DOIUrl":null,"url":null,"abstract":"Self-training methods have proven to be effective in exploiting abundant\nunlabeled data in semi-supervised learning, particularly when labeled data is\nscarce. While many of these approaches rely on a cross-entropy loss function\n(CE), recent advances have shown that the supervised contrastive loss function\n(SupCon) can be more effective. Additionally, unsupervised contrastive learning\napproaches have also been shown to capture high quality data representations in\nthe unsupervised setting. To benefit from these advantages in a semi-supervised\nsetting, we propose a general framework to enhance self-training methods, which\nreplaces all instances of CE losses with a unique contrastive loss. By using\nclass prototypes, which are a set of class-wise trainable parameters, we\nrecover the probability distributions of the CE setting and show a theoretical\nequivalence with it. Our framework, when applied to popular self-training\nmethods, results in significant performance improvements across three different\ndatasets with a limited number of labeled data. Additionally, we demonstrate\nfurther improvements in convergence speed, transfer ability, and hyperparameter\nstability. The code is available at\n\\url{https://github.com/AurelienGauffre/semisupcon/}.","PeriodicalId":501301,"journal":{"name":"arXiv - CS - Machine Learning","volume":null,"pages":null},"PeriodicalIF":0.0000,"publicationDate":"2024-09-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"arXiv - CS - Machine Learning","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/arxiv-2409.07292","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 0

Abstract

Self-training methods have proven to be effective in exploiting abundant unlabeled data in semi-supervised learning, particularly when labeled data is scarce. While many of these approaches rely on a cross-entropy loss function (CE), recent advances have shown that the supervised contrastive loss function (SupCon) can be more effective. Additionally, unsupervised contrastive learning approaches have also been shown to capture high quality data representations in the unsupervised setting. To benefit from these advantages in a semi-supervised setting, we propose a general framework to enhance self-training methods, which replaces all instances of CE losses with a unique contrastive loss. By using class prototypes, which are a set of class-wise trainable parameters, we recover the probability distributions of the CE setting and show a theoretical equivalence with it. Our framework, when applied to popular self-training methods, results in significant performance improvements across three different datasets with a limited number of labeled data. Additionally, we demonstrate further improvements in convergence speed, transfer ability, and hyperparameter stability. The code is available at \url{https://github.com/AurelienGauffre/semisupcon/}.
用于自我训练的统一对比损失
事实证明,在半监督学习中,自我训练方法可以有效利用丰富的无标记数据,尤其是在标记数据稀缺的情况下。虽然这些方法中很多都依赖于交叉熵损失函数(CE),但最近的进展表明,有监督的对比损失函数(SupCon)可能更有效。此外,无监督对比学习方法也被证明可以在无监督环境下捕捉到高质量的数据表示。为了在半监督环境中受益于这些优势,我们提出了一个通用框架来增强自我训练方法,用独特的对比损失来替代所有的 CE 损失实例。通过使用类原型(即一组可训练的类参数),我们覆盖了 CE 设置的概率分布,并展示了与它的理论等价性。当我们的框架应用于流行的自训练方法时,在标注数据数量有限的三个不同数据集上,性能得到了显著提高。此外,我们还证明了收敛速度、转移能力和超参数稳定性的进一步提高。代码可在(url{https://github.com/AurelienGauffre/semisupcon/}.
本文章由计算机程序翻译,如有差异,请以英文原文为准。
求助全文
约1分钟内获得全文 求助全文
来源期刊
自引率
0.00%
发文量
0
×
引用
GB/T 7714-2015
复制
MLA
复制
APA
复制
导出至
BibTeX EndNote RefMan NoteFirst NoteExpress
×
提示
您的信息不完整,为了账户安全,请先补充。
现在去补充
×
提示
您因"违规操作"
具体请查看互助需知
我知道了
×
提示
确定
请完成安全验证×
copy
已复制链接
快去分享给好友吧!
我知道了
右上角分享
点击右上角分享
0
联系我们:info@booksci.cn Book学术提供免费学术资源搜索服务,方便国内外学者检索中英文文献。致力于提供最便捷和优质的服务体验。 Copyright © 2023 布克学术 All rights reserved.
京ICP备2023020795号-1
ghs 京公网安备 11010802042870号
Book学术文献互助
Book学术文献互助群
群 号:481959085
Book学术官方微信