Research map / Architectures and efficiency / Neural architecture search
CATCH: Context-based Meta Reinforcement Learning for Transferrable Architecture Search
Xin Chen, Yawen Duan, Zewei Chen, Hang Xu, Zihao Chen, Xiaodan Liang, Tong Zhang, Zhenguo Li
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 |
|---|---|---|---|
| add meta-training before search | CATCH · NAS-Bench-201 | accuracy | Improved |
| add meta-training before search (CATCH-sfs vs CATCH-meta) | CATCH · NAS-Bench-201 CIFAR-10 | accuracy: best model found in 280/500 trials | Improved |
| search from scratch without meta-training | CATCH-sfs · NAS-Bench-201 | accuracy | No clear effect |
| use CATCH-meta instead of GDAS | one-shot NAS · NAS-Bench-201 CIFAR-100 | accuracy: 70.33 -> 72.57 | Improved |
| use CATCH-meta instead of R-NAS | one-shot NAS · NAS-Bench-201 CIFAR-10 | accuracy: 90.45 -> 91.33 | Improved |
| use CATCH-meta instead of ENAS | one-shot NAS · NAS-Bench-201 ImageNet16-120 | accuracy: 44.92 -> 46.07 | Improved |
| add context encoder (CATCH-zero uses z=0) (ablation) | CATCH-meta · NAS-Bench-201 | search performance | Improved |
| use posterior context rather than random z (CATCH-random) (ablation) | CATCH-meta · NAS-Bench-201 CIFAR-10 and CIFAR-100 | search performance | Improved |
| add context encoder (ablation) | CATCH-meta · NAS-Bench-201 ImageNet16-120 | search performance | No clear effect |
| transfer CATCH-Net-B found on ImageNet to COCO instead of searching on COCO | CATCH-Net-B · COCO | mAP: 42% mAP (vs 43.2 for CATCH-Net-C) | Hurt |
Closest papers by meaning
- Towards Fast Adaptation of Neural Architectures with Meta Learning (ICLR 2020 · similarity 0.90)
- Meta Architecture Search (NeurIPS 2019 · similarity 0.88)
- Rapid Model Architecture Adaption for Meta-Learning (NeurIPS 2022 · similarity 0.88)
- UNAS: Differentiable Architecture Search Meets Reinforcement Learning (CVPR 2020 · similarity 0.88)
- Design Principle Transfer in Neural Architecture Search via Large Language Models (AAAI 2025 · similarity 0.87)
- EG-NAS: Neural Architecture Search with Fast Evolutionary Exploration (AAAI 2024 · similarity 0.86)
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.
