Research map / Learning theory and algorithms / Clustering
Total Variation Graph Neural Networks
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 |
|---|---|---|---|
| Replace MinCutPool/SC-style loss with the GTV-based balanced asymmetric Cheeger loss (TVGNN loss) (ablation) | GNN vertex clustering (GTVConv backbone) · Cora, Citeseer, Pubmed, DBLP (avg) | NMI and ACC: NMI 0.267 -> 0.352; ACC 49.8 -> 60.7 vs ABL1 | Helped |
| Replace GTVConv layers with standard GCN layers (LQV smoothing) (ablation) | TVGNN vertex clustering · Cora, Citeseer, Pubmed, DBLP (avg) | NMI and ACC: NMI 0.352 -> 0.282; ACC 60.7 -> 52.2 | Hurt |
| Use TVGNN (GTVConv + GTV loss) instead of prior GNN clustering methods | Vertex clustering · Cora, Citeseer, Pubmed, DBLP | NMI and ACC average: avg ACC 60.7 vs 53.6 for MinCutPool, the next best | Helped |
| Use GTVConv + GTV loss (TVGNN) for graph pooling in graph classification | Graph classification with hierarchical pooling · 9 graph classification datasets | accuracy: highest mean accuracy on all datasets; e.g., REDDIT-BINARY 86.5 vs 82.6 (DMoN) | Helped |
| Use TVGNN pooling on Pubmed clustering | Vertex clustering · Pubmed | NMI: 0.216 vs VGAE 0.245 (best) | No clear effect |
| Use TVGNN on MUTAG graph classification | Graph classification · MUTAG | accuracy: 88.4 vs 86.7 (DMoN), p=.606 | No clear effect |
| Use TVGNN on Proteins, D&D, COLLAB graph classification | Graph classification · Proteins, D&D, COLLAB | accuracy: differences vs second best not significant (p=.302, .323, .476) | No clear effect |
| Stacked GTVConv with degree-based single Gamma (L_hat) instead of per-feature Gamma^f | GTV feature aggregation · Cora, 32-dim random features (appendix) | GTV of features decrease / convergence | Helped |
Closest papers by meaning
- Spectral Clustering with Graph Neural Networks for Graph Pooling (ICML 2020 · similarity 0.91)
- Do Graph Neural Networks Learn Generalizable Algorithms for Clustering Graphs? (NeurIPS 2026 · similarity 0.90)
- Restructuring Graph for Higher Homophily via Adaptive Spectral Clustering (AAAI 2023 · similarity 0.87)
- Geometry-Aware Edge Pooling for Graph Neural Networks (NeurIPS 2025 · similarity 0.87)
- A Manifold Perspective on the Statistical Generalization of Graph Neural Networks (ICML 2025 · similarity 0.86)
- Graph Neural Networks Are More Than Filters: Revisiting and Benchmarking from A Spectral Perspective (ICLR 2025 · similarity 0.86)
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.