Research map / Graph learning / Graph self-supervised learning

PolyGCL: GRAPH CONTRASTIVE LEARNING via Learnable Spectral Polynomial Filters

ICLR 2024 · Graph learning › Graph self-supervised learning · approach: Graph contrastive learning analysis

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 PolyGCL, a graph contrastive learning method that uses learnable spectral polynomial filters (low- and high-pass) to handle both homophilic and heterophilic graphs.

Problem

Existing graph contrastive learning relies on low-pass GNN encoders and homophily-based objectives, producing smooth representations that perform poorly on heterophilic graphs.

Approach

Uses learnable polynomial spectral filters inside a contrastive framework, with separate low-pass and high-pass polynomial components so the model captures both smooth and high-frequency signals in self-supervised graph representation learning.

Experimental setup

Builds on: graph contrastive learning spectral GNN

Closest papers by meaning

• S3GCL: Spectral, Swift, Spatial Graph Contrastive Learning (ICML 2024 · similarity 0.94)
• Unifying Homophily and Heterophily for Spectral Graph Neural Networks via Triple Filter Ensembles (NeurIPS 2024 · similarity 0.92)
• How Universal Polynomial Bases Enhance Spectral Graph Neural Networks: Heterophily, Over-smoothing, and Over-squashing (ICML 2024 · similarity 0.91)
• Beyond Homophily: Graph Contrastive Learning with Macro-Micro Message Passing (AAAI 2025 · similarity 0.91)
• HeroFilter: Adaptive Spectral Graph Filter for Varying Heterophilic Relations (NeurIPS 2025 · similarity 0.90)
• Does GCL Need a Large Number of Negative Samples? Enhancing Graph Contrastive Learning with Effective and Efficient Negative Sampling (AAAI 2025 · similarity 0.90)