Research map / Graph learning / Graph algorithms and clustering
Arch-Graph: Acyclic Architecture Relation Predictor for Task-Transferable Neural Architecture Search
Check 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 the MWAS graph ordering instead of Arch-Graph-zero insertion-sort ordering (ablation) | Arch-Graph · TransNAS-Bench-101 macro | average rank: 7.83 -> 5.24; improves the rank by 3.14 on average | Helped |
| Use the MWAS graph ordering instead of Arch-Graph-zero insertion-sort ordering (ablation) | Arch-Graph · TransNAS-Bench-101 micro | average rank: 14.7 -> 12.2; improves the rank by 5.28 on average | Helped |
| Use Task2Vec task embeddings rather than randomly initialized task vectors (ablation) | Arch-Graph · TransNAS-Bench-101, six target tasks | average rank: 24.13 -> 5.24 | Helped |
| Use Task2Vec task embeddings rather than randomly initialized task vectors (ablation) | Arch-Graph · TransNAS-Bench-101, six target tasks | variance of average rank: 692.03 -> 0.63 | Helped |
| Transfer the predictor from a source task instead of pretraining and finetuning on the same task (ablation) | Arch-Graph · TransNAS-Bench-101 macro | average rank: 12.68 -> 5.24 | Helped |
| Transfer the predictor from a source task instead of pretraining and finetuning on the same task (ablation) | Arch-Graph · TransNAS-Bench-101 micro | average rank: 22.15 -> 12.2 | Helped |
| Use Arch-Graph rather than weakNAS | NAS search on CIFAR-100 · NAS-Bench-201 CIFAR-100 | accuracy: 73.42 -> 73.38 | Hurt |
| Use Arch-Graph rather than weakNAS | NAS search on CIFAR-100 · NAS-Bench-201 CIFAR-100 | Kendall rank coefficient: 0.49 -> 0.67 | Helped |
| Use Arch-Graph rather than weakNAS-t | Transfer NAS · TransNAS-Bench-101 macro | average rank: 15.43 -> 5.24 | Helped |
| Use Arch-Graph rather than weakNAS-t | Transfer NAS · TransNAS-Bench-101 micro | average rank: 35.73 -> 12.2 | Helped |
Closest papers by meaning
- Multi-task Graph Neural Architecture Search with Task-aware Collaboration and Curriculum (NeurIPS 2023 · similarity 0.91)
- TransNAS-Bench-101: Improving Transferability and Generalizability of Cross-Task Neural Architecture Search (CVPR 2021 · similarity 0.90)
- A Semi-Supervised Assessor of Neural Architectures (CVPR 2020 · similarity 0.89)
- Aux-NAS: Exploiting Auxiliary Labels with Negligibly Extra Inference Cost (ICLR 2024 · similarity 0.87)
- Encodings for Prediction-based Neural Architecture Search (ICML 2024 · similarity 0.87)
- Disentangled Continual Graph Neural Architecture Search with Invariant Modular Supernet (ICML 2024 · similarity 0.87)
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.