Research map / Architectures and efficiency / Neural architecture search
Generic Neural Architecture Search via Regression
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 |
|---|---|---|---|
| Search the synthetic-signal proxy task using a small set of task-specific architecture labels | GenNAS · NASBench-101, 20 sampled architectures | Spearman's ρ: 0.82±0.01 -> 0.86±0.02 | Helped |
| Combine signal tasks across three intermediate stages instead of using one stage (ablation) | GenNAS · NASBench-101, CIFAR-10 | Spearman's ρ: 0.81 -> 0.85 | Helped |
| Use medium- or high-frequency Sin1D/Sin2D rather than low-frequency signals (ablation) | GenNAS synthetic-signal proxy · NASBench-101, CIFAR-10; stages S1 and S2 | Spearman's ρ: 0.13 -> 0.64 | Helped |
| Use weakly supervised regression instead of classification with the same training samples (ablation) | GenNAS · NASBench-101 | Spearman's ρ: around 0.3 -> around 0.85 | Helped |
| Increase batch size from 1 to 16 (ablation) | GenNAS · NASBench-101 | ranking performance | Helped |
| Increase batch size beyond 16 (ablation) | GenNAS · NASBench-101 | ranking performance | Hurt |
| Use 100 training iterations (ablation) | GenNAS · NASBench-101 | ranking performance | Helped |
| Use zero-initialized bias instead of the default bias initialization (ablation) | GenNAS · NASBench-101 | Spearman's ρ: 0.841±0.021 -> 0.856±0.021 | Helped |
| Use Kaiming weight initialization instead of default initialization (ablation) | GenNAS · NASBench-101 | Spearman's ρ: 0.841±0.021 -> 0.830±0.032 | Hurt |
| Transfer the searched NASBench-101 proxy task to NASBench-201 | GenNAS · NASBench-201, CIFAR-10 | Spearman's ρ: 0.87 -> 0.90 | Helped |
| Use task-agnostic synthetic-signal regression rather than a classification proxy | GenNAS · NDS DARTS, CIFAR-10 | Spearman's ρ: 0.63 -> 0.86 | Helped |
Closest papers by meaning
- TransNAS-Bench-101: Improving Transferability and Generalizability of Cross-Task Neural Architecture Search (CVPR 2021 · similarity 0.93)
- NAS-Bench-201: Extending the Scope of Reproducible Neural Architecture Search (ICLR 2020 · similarity 0.91)
- Neural Architecture Search on ImageNet in Four GPU Hours: A Theoretically Inspired Perspective (ICLR 2021 · similarity 0.91)
- Towards Neural Architecture Search through Hierarchical Generative Modeling (ICML 2024 · similarity 0.90)
- Insights from the Use of Previously Unseen Neural Architecture Search Datasets (CVPR 2024 · similarity 0.90)
- Efficient Few-Shot Neural Architecture Search by Counting the Number of Nonlinear Functions (AAAI 2025 · 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.