Research map / Architectures and efficiency / Neural architecture search
Graph HyperNetworks for Neural Architecture Search
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 graph-topology-conditioned weight generation instead of a one-shot shared-parameter model | GHN · CIFAR-10, Top-50 | Pearson correlation: 0.31 -> 0.48 | Helped |
| use GHN-generated weights to rank candidate architectures | random-search NAS on CIFAR-10 · CIFAR-10 | search cost (GPU days): 0.42 | Helped |
| use forward-backward rather than synchronous message passing (ablation) | GHN · CIFAR-10 | predicted-performance correlation | Helped |
| increase the number of forward-backward propagation steps (ablation) | GHN · CIFAR-10 | predicted-performance correlation | Helped |
| train the GHN on smaller graphs rather than larger graphs (ablation) | GHN · CIFAR-10 | predicted-performance correlation | Helped |
| pass graph embeddings between stacked modules (ablation) | stacked GHN · CIFAR-10 | Pearson correlation: 0.24 -> 0.44 (Random-100); 0.15 -> 0.37 (Top-50) | Helped |
| share parameters across stacked modules (ablation) | stacked GHN with embedding passing · CIFAR-10 | Pearson correlation: 0.44 -> 0.68 (Random-100); 0.37 -> 0.48 (Top-50) | Helped |
Closest papers by meaning
- Do Not Train It: A Linear Neural Architecture Search of Graph Neural Networks (ICML 2023 · similarity 0.92)
- Rethinking Graph Neural Architecture Search From Message-Passing (CVPR 2021 · similarity 0.91)
- NAS-Bench-Graph: Benchmarking Graph Neural Architecture Search (NeurIPS 2022 · similarity 0.91)
- Graph Differentiable Architecture Search with Structure Learning (NeurIPS 2021 · similarity 0.90)
- Large-Scale Graph Neural Architecture Search (ICML 2022 · similarity 0.90)
- Hypergraph Neural Architecture Search (AAAI 2024 · similarity 0.89)
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.