EDiS: Edge Disjoint Subgraph Sparsification Framework for Graph Neural Networks
EDiS sparsifies GNN training by caching edge-disjoint subgraphs and recomposing them each epoch under an edge budget.
EDiS separates one-time structural extraction from per-epoch graph composition for sparse graph neural network training. It decomposes a graph once into cacheable edge-disjoint subgraphs, then recombines them under edge-budget and retention constraints without re-extracting structure. The default selector uses feature-based scores and successive maximum-score covering forests, and the authors show it preserves high-score cut edges. Across 19 node-classification benchmarks against 17 baselines at the same edge budget, EDiS recorded the highest mean accuracy or ROC-AUC and the lowest average rank.
- Decomposes a graph once into cacheable edge-disjoint subgraphs
- Recombines those subgraphs each epoch under a fixed edge budget
- Default selector uses feature scores and covering forests
- Highest mean score and lowest rank versus 17 baselines on 19 benchmarks
- Clearest gains from decomposition and epoch variation at tight budgets
From the original article
1 min read at huggingface.co
Sparse GNN training reduces computation, but deciding which edges to keep can be costly. Reusing one sparse graph is cheap, but locks training to a fixed topology, while varying it across epochs can require repeated sampling or recomputation. We introduce EDiS (Edge-Disjoint Subgraph sparsification framework), which separates one-time structural extraction from per-epoch graph composition.
By Sai Karthik Navuluru, Siddhartha Shankar Das, Franck Dernoncourt · huggingface.co
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.