Research map / Learning theory and algorithms / Optimal transport
Optimal Transport Kernels for Sequential and Parallel Neural Architecture Search
Vu Nguyen, Tam Le, Makoto Yamada, Michael A. Osborne
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 |
|---|---|---|---|
| Replace standard OT (indefinite) distance with tree-Wasserstein (negative definite) distance for GP kernel in sequential NAS (ablation) | BO for NAS (GP surrogate) · NB101/NB201 sequential NAS | best accuracy vs iterations | Improved |
| Use 2-gram operation representation instead of 1-gram in tree-Wasserstein (ablation) | TW distance for NAS · NB101/NB201 sequential NAS | best accuracy vs iterations | Improved |
| Use evolutionary candidate pool generation instead of random generation (ablation) | Candidate pool P_t for BO · NAS benchmarks | best accuracy | Improved |
| Use k-DPP with quality (GP-based) for parallel batch selection instead of GP-BUCB/KB, Thompson sampling, and vanilla k-DPP | Parallel NAS batch selection · NB101 parallel NAS (budget 500, batch 5) | best accuracy vs iterations | Improved |
| Use TW-2G distance instead of path distance within k-DPP quality (ablation) | Parallel NAS k-DPP quality · NB201 (ImageNet, CIFAR-100) | best accuracy | Improved |
| Replace OT distance with Gromov-Wasserstein or path encoding within BO (ablation) | BO distances for NAS · NAS benchmarks | best accuracy | Hurt |
| Use GP-UCB acquisition instead of expected improvement (ablation) | Sequential BO · NAS benchmarks | best accuracy | Improved |
| Add quality term to k-DPP (vs diversity only) (ablation) | Parallel NAS k-DPP · NB101 parallel NAS | best accuracy | Improved |
Closest papers by meaning
- Neural Architecture Generator Optimization (NeurIPS 2020 · similarity 0.89)
- Interpretable Neural Architecture Search via Bayesian Optimisation with Weisfeiler-Lehman Kernels (ICLR 2021 · similarity 0.89)
- Evaluating The Search Phase of Neural Architecture Search (ICLR 2020 · similarity 0.89)
- NAS-Bench-201: Extending the Scope of Reproducible Neural Architecture Search (ICLR 2020 · similarity 0.89)
- Understanding and Accelerating Neural Architecture Search With Training-Free and Theory-Grounded Metrics (TPAMI 2023 · similarity 0.89)
- Efficient Few-Shot Neural Architecture Search by Counting the Number of Nonlinear Functions (AAAI 2025 · similarity 0.88)
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.
