Research map / Graph learning / GNN architectures
Graph Partitioning and Sparse Matrix Ordering using Reinforcement Learning and Graph Neural Networks
Alice Gatti, Zhixiong Hu, Tess Smidt, Esmond G. Ng, Pieter Ghysels
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 DRL partitioning of the coarsest graph instead of METIS at the coarsest level (DRL vs DRL_METIS) | multilevel DRL edge partitioner · Delaunay test sets | normalized cut | No clear effect |
| train on Delaunay graphs and test on unseen finite element triangulations (GradedL, Hole3, Hole6) | DRL edge partitioner · FE triangulations | normalized cut, cut, balance | No clear effect |
| train a separate agent on SuiteSparse 2D/3D discretizations | DRL edge partitioner · SuiteSparse 5,000 to 90,000 nodes | cut and normalized cut | Improved |
| use DRL vertex separator with METIS at the coarsest level instead of METIS alone | DRL vertex separator · Delaunay and Hole6 sets | normalized separator | Mixed |
| use DRL vertex separator on SuiteSparse (101 matrices) | DRL vertex separator · SuiteSparse 5,000 to 90,000 vertices | normalized separator | Hurt |
| use DRL_ND nested dissection ordering instead of COLAMD | SuperLU sparse factorization · Delaunay graphs (100) and SuiteSparse (101) | fill in L and U | Improved |
Closest papers by meaning
- Combinatorial Optimization with Graph Convolutional Networks and Guided Tree Search (NeurIPS 2018 · similarity 0.89)
- Exact Combinatorial Optimization with Graph Convolutional Neural Networks (NeurIPS 2019 · similarity 0.87)
- All in a Row: Compressed Convolution Networks for Graphs (ICML 2023 · similarity 0.87)
- A Structure-Aware Framework for Learning Device Placements on Computation Graphs (NeurIPS 2024 · similarity 0.85)
- Two Heads Are Better Than One: Boosting Graph Sparse Training via Semantic and Topological Awareness (ICML 2024 · similarity 0.85)
- Generalizing CNNs to graphs with learnable neighborhood quantization (NeurIPS 2024 · similarity 0.85)
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.
