Research map / Label-efficient and robust learning / Self-supervised and contrastive
Zero-Mean Regularized Spectral Contrastive Learning: Implicitly Mitigating Wrong Connections in Positive-Pair Graphs
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's abstract only. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Proposes a zero-mean regularized spectral contrastive loss that implicitly mitigates wrong connections in positive-pair graphs for self-supervised representation learning.
Problem
Spectral contrastive learning relies on positive-pair graphs, which may contain wrong connections (positive pairs from different underlying classes), harming representation quality.
Approach
Adds a zero-mean regularization to the spectral contrastive loss so that the learned representations are implicitly robust to wrong edges in the positive-pair graph, without explicitly identifying them.
Experimental setup
Builds on: spectral contrastive loss
Closest papers by meaning
- Does GCL Need a Large Number of Negative Samples? Enhancing Graph Contrastive Learning with Effective and Efficient Negative Sampling (AAAI 2025 · similarity 0.89)
- Perfect Alignment May be Poisonous to Graph Contrastive Learning (ICML 2024 · similarity 0.87)
- A New Mechanism for Eliminating Implicit Conflict in Graph Contrastive Learning (AAAI 2024 · similarity 0.87)
- On the Similarities of Embeddings in Contrastive Learning (ICML 2025 · similarity 0.87)
- Diversity Is All You Need for Contrastive Learning: Spectral Bounds on Gradient Magnitudes (NeurIPS 2025 · similarity 0.87)
- Non-negative Contrastive Learning (ICLR 2024 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.