paper-with-me

홈 › Papers

A Minimal Model of Representation Collapse: Frustration, Stop-Gradient, and Dynamics

2026-04-11 · Louie Hong Yao, Yuhao Li, Shengchao Liu arxiv

Self-supervised representation learning is central to modern machine learning because it extracts structured latent features from unlabeled data and enables robust transfer across tasks and domains. However, it can suffer from representation collapse, a widely observed failure mode in which embeddings lose discriminative structure and distinct inputs become indistinguishable. To understand the mechanisms that drive collapse and the ingredients that prevent it, we introduce a minimal embedding-only model whose gradient-flow dynamics and fixed points can be analyzed in closed form, using a classification-representation setting as a concrete playground where collapse is directly quantified through the contraction of label-embedding geometry. We illustrate that the model does not collapse when the data are perfectly classifiable, while a small fraction of frustrated samples that cannot be classified consistently induces collapse through an additional slow time scale that follows the early performance gain. Within the same framework, we examine collapse prevention by adding a shared projection head and applying stop-gradient at the level of the training dynamics. We analyze the resulting fixed points and develop a dynamical mean-field style self-consistency description, showing that stop-gradient enables non-collapsed solutions and stabilizes finite class separation under frustration. We further verify empirically that the same qualitative dynamics and collapse-prevention effects appear in a linear teacher-student model, indicating that the minimal theory captures features that persist beyond the pure embedding setting.

📄 PDF Abstract BibTeX arXiv:2604.09979

Code (0)

등록된 구현이 없습니다.

Tasks

Representation Learning

Similar Papers 제목 키워드 기반

Implicit Contrastive Representation Learning with Guided Stop-gradient

2025-03-12 · NeurIPS 2023 11 · Byeongchan Lee, Sehyun Lee

In self-supervised representation learning, Siamese networks are a natural architecture for learning transformation-invariance by bringing representations of positive pairs closer together. But it is prone to collapse in…

Contrastive LearningRepresentation Learning

Understanding self-supervised Learning Dynamics without Contrastive Pairs

2021-02-12 · Yuandong Tian, Xinlei Chen, Surya Ganguli

While contrastive approaches of self-supervised learning (SSL) learn representations by minimizing the distance between two augmented views of the same data point (positive pairs) and maximizing views from different data…

Self-Supervised Learning

The Triad of Failure Modes and a Possible Way Out

2023-09-27 · Emanuele Sansone

We present a novel objective function for cluster-based self-supervised learning (SSL) that is designed to circumvent the triad of failure modes, namely representation collapse, cluster collapse, and the problem of invar…

Self-Supervised Learning

Identifying Human Needs through Social Media: A study on Indian cities during COVID-19

2022-07-01 · NAACL (SocialNLP) 2022 7 · Sunny Rai, Rohan Joseph, Prakruti Singh Thakur, Mohammed Abdul Khaliq

In this paper, we present a minimally-supervised approach to identify human needs expressed in tweets. Taking inspiration from Frustration-Aggression theory, we trained RoBERTa model to classify tweets expressing frustra…

Language ModelingLanguage Modelling

How Does SimSiam Avoid Collapse Without Negative Samples? A Unified Understanding with Self-supervised Contrastive Learning

2022-03-30 · Chaoning Zhang, Kang Zhang, Chenshuang Zhang, Trung X. Pham 외

To avoid collapse in self-supervised learning (SSL), a contrastive loss is widely used but often requires a large number of negative samples. Without negative samples yet achieving competitive performance, a recent work …

Contrastive LearningSelf-Supervised Learning