Research map / Optimization / Optimizers
Adaptive Stochastic Natural Gradient Method for One-Shot 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 ASNG-NAS rather than the human-designed CE, SII, or BASE models | CelebA inpainting · Center, Pixel, and Half masks | PSNR [dB]: Center: CE 28.5, SII 19.4, BASE 27.1; ASNG-NAS (Cat) 29.2, ASNG-NAS (Int) 29.3. Pixel: CE 22.9, SII 22.8, BASE 27.5; Cat | Helped |
| Use ASNG-NAS rather than the human-designed CE, SII, or BASE models | CelebA inpainting · Center, Pixel, and Half masks | SSIM: Center: CE 0.912, SII 0.907, BASE 0.883; Cat 0.903, Int 0.911. Pixel: CE 0.730, SII 0.710, BASE 0.836; Cat 0.905, Int 0. | Mixed |
| Use ASNG-NAS rather than E-CAE | CelebA inpainting · Center, Pixel, and Half masks | PSNR [dB]: Center: 29.9 vs 29.2 (Cat) and 29.3 (Int); Pixel: 27.8 vs 28.4 (Cat) and 28.6 (Int); Half: 21.1 vs 20.5 (Cat) and 20.6 ( | Mixed |
| Use ASNG-NAS rather than E-CAE | CelebA inpainting · Center, Pixel, and Half masks | SSIM: Center: 0.934 vs 0.903 (Cat) and 0.911 (Int); Pixel: 0.887 vs 0.905 (Cat) and 0.909 (Int); Half: 0.771 vs 0.779 (Cat) an | Mixed |
| Represent kernel and channel sizes as integer variables instead of categorical choices (ablation) | ASNG-NAS on CelebA inpainting · Pixel mask | PSNR [dB]: 28.4 -> 28.6 | Helped |
| Represent kernel and channel sizes as integer variables instead of categorical choices (ablation) | ASNG-NAS on CelebA inpainting · Center, Pixel, and Half masks | SSIM: Center: 0.903 -> 0.911; Pixel: 0.905 -> 0.909; Half: 0.779 -> 0.779 | Mixed |
| Represent kernel and channel sizes as integer variables instead of categorical choices (ablation) | ASNG-NAS on CelebA inpainting · Center and Half masks | PSNR [dB]: Center: 29.2 -> 29.3; Half: 20.5 -> 20.6 | Helped |
Closest papers by meaning
- SNAS: stochastic neural architecture search (ICLR 2019 · similarity 0.91)
- Geometry-Aware Gradient Algorithms for Neural Architecture Search (ICLR 2021 · similarity 0.91)
- iDARTS: Differentiable Architecture Search with Stochastic Implicit Gradients (ICML 2021 · similarity 0.91)
- XNAS: Neural Architecture Search with Expert Advice (NeurIPS 2019 · similarity 0.91)
- Efficient Few-Shot Neural Architecture Search by Counting the Number of Nonlinear Functions (AAAI 2025 · similarity 0.90)
- EG-NAS: Neural Architecture Search with Fast Evolutionary Exploration (AAAI 2024 · similarity 0.90)
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.