Understanding encoder–decoder structures in machine learning using information measures

IF 3.4 2区 工程技术 Q2 ENGINEERING, ELECTRICAL & ELECTRONIC
Jorge F. Silva , Victor Faraggi , Camilo Ramirez , Alvaro Egaña , Eduardo Pavez
{"title":"Understanding encoder–decoder structures in machine learning using information measures","authors":"Jorge F. Silva ,&nbsp;Victor Faraggi ,&nbsp;Camilo Ramirez ,&nbsp;Alvaro Egaña ,&nbsp;Eduardo Pavez","doi":"10.1016/j.sigpro.2025.109983","DOIUrl":null,"url":null,"abstract":"<div><div>We present a theory of representation learning to model and understand the role of encoder–decoder design in machine learning (ML) from an information-theoretic angle. We use two main information concepts, information sufficiency (IS) and mutual information loss to represent predictive structures in machine learning. Our first main result provides a functional expression that characterizes the class of probabilistic models consistent with an IS encoder–decoder latent predictive structure. This result formally justifies the encoder–decoder forward stages many modern ML architectures adopt to learn latent (compressed) representations for classification. To illustrate IS as a realistic and relevant model assumption, we revisit some known ML concepts and present some interesting new examples: invariant, robust, sparse, and digital models. Furthermore, our IS characterization allows us to tackle the fundamental question of how much performance could be lost, using the cross entropy risk, when a given encoder–decoder architecture is adopted in a learning setting. Here, our second main result shows that a mutual information loss quantifies the lack of expressiveness attributed to the choice of a (biased) encoder–decoder ML design. Finally, we address the problem of universal cross-entropy learning with an encoder–decoder design where necessary and sufficiency conditions are established to meet this requirement. In all these results, Shannon’s information measures offer new interpretations and explanations for representation learning.</div></div>","PeriodicalId":49523,"journal":{"name":"Signal Processing","volume":"234 ","pages":"Article 109983"},"PeriodicalIF":3.4000,"publicationDate":"2025-03-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Signal Processing","FirstCategoryId":"5","ListUrlMain":"https://www.sciencedirect.com/science/article/pii/S0165168425000970","RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q2","JCRName":"ENGINEERING, ELECTRICAL & ELECTRONIC","Score":null,"Total":0}
引用次数: 0

Abstract

We present a theory of representation learning to model and understand the role of encoder–decoder design in machine learning (ML) from an information-theoretic angle. We use two main information concepts, information sufficiency (IS) and mutual information loss to represent predictive structures in machine learning. Our first main result provides a functional expression that characterizes the class of probabilistic models consistent with an IS encoder–decoder latent predictive structure. This result formally justifies the encoder–decoder forward stages many modern ML architectures adopt to learn latent (compressed) representations for classification. To illustrate IS as a realistic and relevant model assumption, we revisit some known ML concepts and present some interesting new examples: invariant, robust, sparse, and digital models. Furthermore, our IS characterization allows us to tackle the fundamental question of how much performance could be lost, using the cross entropy risk, when a given encoder–decoder architecture is adopted in a learning setting. Here, our second main result shows that a mutual information loss quantifies the lack of expressiveness attributed to the choice of a (biased) encoder–decoder ML design. Finally, we address the problem of universal cross-entropy learning with an encoder–decoder design where necessary and sufficiency conditions are established to meet this requirement. In all these results, Shannon’s information measures offer new interpretations and explanations for representation learning.
求助全文
约1分钟内获得全文 求助全文
来源期刊
Signal Processing
Signal Processing 工程技术-工程:电子与电气
CiteScore
9.20
自引率
9.10%
发文量
309
审稿时长
41 days
期刊介绍: Signal Processing incorporates all aspects of the theory and practice of signal processing. It features original research work, tutorial and review articles, and accounts of practical developments. It is intended for a rapid dissemination of knowledge and experience to engineers and scientists working in the research, development or practical application of signal processing. Subject areas covered by the journal include: Signal Theory; Stochastic Processes; Detection and Estimation; Spectral Analysis; Filtering; Signal Processing Systems; Software Developments; Image Processing; Pattern Recognition; Optical Signal Processing; Digital Signal Processing; Multi-dimensional Signal Processing; Communication Signal Processing; Biomedical Signal Processing; Geophysical and Astrophysical Signal Processing; Earth Resources Signal Processing; Acoustic and Vibration Signal Processing; Data Processing; Remote Sensing; Signal Processing Technology; Radar Signal Processing; Sonar Signal Processing; Industrial Applications; New Applications.
×
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