arXivDaily arXiv每日学术速递 周一至周五更新
​通知:联邦节假日调休:哥伦布日 / 原住民日(Columbus Day / Indigenous Peoples' Day),今日(13日)无更新。
arXiv 2605.03517cs.LGstat.ML

通过潜在分布匹配理解自监督学习

Understanding Self-Supervised Learning via Latent Distribution Matching

Friedrich Miescher Institute for Biomedical Research, 4056 Basel, Switzerland(弗里德里希·迈斯彻生物医学研究所);Faculty of Science, University of Basel, 4033 Basel, Switzerland(巴塞尔大学科学学院)

Fabian A Mikulasch, Friedemann Zenke

AI总结:

本文将自监督学习形式化为潜在分布匹配(LDM),通过对齐和均匀性最大化潜在表示的对数概率和熵,统一了多种SSL方法,并推导出用于高维时间序列的非线性无采样贝叶斯滤波模型。

摘要:

自监督学习(SSL)擅长从复杂数据中学习通用潜在表示,但缺乏统一的理论框架来解释现有各种方法并指导新方法的设计。我们将SSL视为潜在分布匹配(LDM):学习表示以最大化其在假设潜在模型下的对数概率(对齐),同时最大化潜在熵以防止坍塌(均匀性)。这一观点将独立成分分析与对比、非对比和预测性SSL方法(包括停止梯度方法)统一起来。利用LDM,我们推导出一个非线性的、无采样的贝叶斯滤波模型,其中包含基于卡尔曼的预测器,用于高维时间序列。我们进一步证明,在温和假设下,即使使用非线性预测器,预测性LDM也能产生可识别的潜在表示。总体而言,LDM阐明了现有SSL方法背后的假设,并为开发新方法提供了原则性指导。

英文摘要:

Self-supervised learning (SSL) excels at finding general-purpose latent representations from complex data, yet lacks a unifying theoretical framework that explains the diverse existing methods and guides the design of new ones. We cast SSL as latent distribution matching (LDM): learning representations that maximize their log-probability under an assumed latent model (alignment), while maximizing latent entropy to prevent collapse (uniformity). This view unifies independent component analysis with contrastive, non-contrastive, and predictive SSL methods, including stop gradient approaches. Leveraging LDM, we derive a nonlinear, sampling-free Bayesian filtering model with a Kalman-based predictor for high-dimensional timeseries. We further prove that predictive LDM yields identifiable latent representations under mild assumptions, even with nonlinear predictors. Overall, LDM clarifies the assumptions behind established SSL methods and provides principled guidance for developing new approaches.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑