Research map / Graph learning / Graph self-supervised learning
Beyond Smoothing: Unsupervised Graph Representation Learning with Edge Heterophily Discriminating
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use GREET instead of the strongest listed baseline | unsupervised node representation learning · heterophilic benchmarks (Chameleon, Squirrel, Actor, Cornell, Texas) | accuracy: 63.44 -> 63.64 (Chameleon); 41.17 -> 42.29 (Squirrel); 35.86 -> 36.55 (Actor); 82.16 -> 85.14 (Cornell); 84.86 -> 87.03 | Helped |
| Use GREET instead of the strongest listed baseline | unsupervised node representation learning · homophilic benchmarks (PubMed, Amazon Computer, Amazon Photo) | accuracy: 80.37 -> 80.29 (PubMed); 88.92 -> 87.94 (Amazon Computer); 93.24 -> 92.85 (Amazon Photo) | Hurt |
| Add an edge discriminator (ablation) | GREET · Cornell | accuracy: 82.16 -> 85.14 | Helped |
| Add dual-channel encoding (ablation) | GREET · Cornell | accuracy: 77.57 -> 85.14 | Helped |
| Use structural encodings instead of a GNN-based discriminator (ablation) | GREET · Cornell | accuracy: 82.97 -> 85.14 | Helped |
| Use pivot-anchored ranking instead of ranking homophilic against heterophilic edges (ablation) | GREET · Cornell | accuracy: 79.19 -> 85.14 | Helped |
| Use robust dual-channel contrastive loss instead of InfoNCE (ablation) | GREET · Cornell | accuracy: 77.57 -> 85.14 | Helped |
| Add the heterophilic-view representation alongside the homophilic-view representation (ablation) | GREET · Cornell | accuracy: 75.95 -> 85.14 | Helped |
| Add the homophilic-view representation alongside the heterophilic-view representation (ablation) | GREET · Cornell | accuracy: 82.97 -> 85.14 | Helped |
Closest papers by meaning
- Multiplex Graph Representation Learning with Homophily and Consistency (AAAI 2025 · similarity 0.90)
- Unraveling the Impact of Heterophilic Structures on Graph Positive-Unlabeled Learning (ICML 2024 · similarity 0.90)
- PC-Conv: Unifying Homophily and Heterophily with Two-Fold Filtering (AAAI 2024 · similarity 0.89)
- Integrating Co-Training with Edge Discrimination to Enhance Graph Neural Networks Under Heterophily (AAAI 2025 · similarity 0.89)
- Transitivity-Preserving Graph Representation Learning for Bridging Local Connectivity and Role-Based Similarity (AAAI 2024 · similarity 0.88)
- Beyond Redundancy: Information-aware Unsupervised Multiplex Graph Structure Learning (NeurIPS 2024 · similarity 0.88)
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.