Research map / Architectures and efficiency / Neural architecture search
Sample-Efficient Neural Architecture Search by Learning Actions for Monte Carlo Tree Search
Linnan Wang, Saining Xie, Teng Li, Rodrigo Fonseca, Yuandong Tian
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 learned latent actions instead of manual actions | MCTS NAS · NASBench-101 | speedup vs random search: 700x faster | Improved |
| use learned latent actions instead of random search | MCTS NAS · ConvNet-60K | speedup vs random search: 22x faster | Improved |
| learn actions (LaNAS) vs second best baseline | NAS search · NASBench-101 | final test error (%): 0.137% -> 0.011% | Improved |
| use global action space instead of sequential action space (ablation) | MCTS · CIFAR-10 plain CNN motivating example | search speed | Improved |
| use linear classifier instead of MLP for splitting (ablation) | LaNAS · NASBench-101 | samples to optimum | Improved |
| increase tree height (ablation) | LaNAS · NASBench-101 | samples to find optimum | Improved |
| decrease number of selects per iteration (ablation) | LaNAS · NASBench-101 | samples to find optimum | Improved |
| add random initialization samples (init > 0) (ablation) | LaNAS · NASBench-101 | search performance | Improved |
| use init=2000 instead of init=100-400 (ablation) | LaNAS · NASBench-101 | time to top 5% | Hurt |
| add exploration term c from 0 to 0.1 in UCB (ablation) | LaNAS · NASBench-101 | search performance | Improved |
| use large c above 0.5 in UCB (ablation) | LaNAS · NASBench-101 | search performance | Hurt |
| use Bayesian sampling instead of random sampling within partition (ablation) | LaNAS · NASBench-101 | final result | Mixed |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- FairDARTS — Fair DARTS: Eliminating Unfair Advantages in Differentiable Architecture Search (ECCV 2020) · better on 2 of 2 shared results
- DARTS — DARTS-: Robustly Stepping out of Performance Collapse Without Indicators (ICLR 2021) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Prioritized Architecture Sampling With Monto-Carlo Tree Search (CVPR 2021 · similarity 0.92)
- Bridging the Gap between Sample-based and One-shot Neural Architecture Search with BONAS (NeurIPS 2020 · similarity 0.90)
- Neural Architecture Search on ImageNet in Four GPU Hours: A Theoretically Inspired Perspective (ICLR 2021 · similarity 0.90)
- Evolving Search Space for Neural Architecture Search (ICCV 2021 · similarity 0.90)
- Efficient Few-Shot Neural Architecture Search by Counting the Number of Nonlinear Functions (AAAI 2025 · similarity 0.89)
- Towards Neural Architecture Search through Hierarchical Generative Modeling (ICML 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.
