Research map / Graph learning / Graph algorithms and clustering
Reinforced Genetic Algorithm Learning for Optimizing Computation Graphs
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 |
|---|---|---|---|
| Predict graph-conditioned proposal distributions for BRKGA with a GNN instead of using uniform proposals | BRKGA · TF Runtime test set | % improvement over BRKGA 5K: 0.00% -> 7.09% | Helped |
| Predict graph-conditioned proposal distributions for BRKGA with a GNN instead of using uniform proposals | BRKGA · TF Peak Memory test set | % improvement over BRKGA 5K: 0.00% -> 3.56% | Helped |
| Predict graph-conditioned proposal distributions for BRKGA with a GNN instead of using uniform proposals | BRKGA · Synthetic Runtime test set | % improvement over BRKGA 5K: 0.00% -> 4.81% | Helped |
| Apply a policy trained on TensorFlow graphs to XLA graphs | REGAL · XLA Runtime | % improvement over BRKGA 5K: 0.58% | Helped |
| Apply a policy trained on TensorFlow graphs to XLA graphs | REGAL · XLA Peak Memory | % improvement over BRKGA 5K: 3.74% | Helped |
| Use placement actions only and omit scheduling actions (ablation) | REGAL · TF Peak Memory test set | % improvement over BRKGA 5K: 3.56% -> -0.2% | Hurt |
| Use scheduling actions only and omit placement actions (ablation) | REGAL · TF Peak Memory test set | % improvement over BRKGA 5K: 0.00% -> 3.65% | Helped |
| Add placement actions to scheduling actions (ablation) | REGAL · TF Peak Memory test set | % improvement over BRKGA 5K: 3.65% -> 3.56% | Hurt |
| Add placement actions to scheduling actions (ablation) | REGAL · TF Peak Memory validation set | % improvement over BRKGA 5K: 4.4% -> 4.67% | Helped |
| Use GNN message passing to incorporate graph structure rather than T=0 (ablation) | REGAL policy · TF Runtime validation | validation reward | Helped |
Closest papers by meaning
- Transferable Graph Optimizers for ML Compilers (NeurIPS 2020 · similarity 0.92)
- DOGE-Train: Discrete Optimization on GPU with End-to-End Training (AAAI 2024 · similarity 0.86)
- EvoOptiGraph: Weakness-Driven Coevolution via Graph-Based Structural Generation for Optimization Modeling (NeurIPS 2026 · similarity 0.85)
- A Structure-Aware Framework for Learning Device Placements on Computation Graphs (NeurIPS 2024 · similarity 0.85)
- OptEx: Expediting First-Order Optimization with Approximately Parallelized Iterations (NeurIPS 2024 · similarity 0.85)
- Symmetric Replay Training: Enhancing Sample Efficiency in Deep Reinforcement Learning for Combinatorial Optimization (ICML 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.