Research map / Architectures and efficiency / Neural architecture search
XNAS: Neural Architecture Search with Expert Advice
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 XNAS-Large instead of SharpDarts | CIFAR-10 architecture search and evaluation · CIFAR-10 test | top-1 test error (%): 1.93 -> 1.60 | Helped |
| Increase the XNAS network from 36 to 50 initial channels (ablation) | XNAS · CIFAR-10 test | top-1 test error (%): 1.81 -> 1.60 | Helped |
| Transfer the CIFAR-10-searched cell to ImageNet | XNAS · ImageNet test | top-1 test error (%): 24.3 -> 24.0 | Helped |
| Transfer the CIFAR-10-searched cell to CIFAR-100 | XNAS-Small · CIFAR-100 test | top-1 test error (%): 10.7 -> 13.6 | Hurt |
| Transfer the CIFAR-10-searched cell to Fashion-MNIST | XNAS-Small · Fashion-MNIST test | top-1 test error (%): 3.65 -> 3.64 | Helped |
| Transfer the CIFAR-10-searched cell to SVHN | XNAS-Small · SVHN test | top-1 test error (%): 1.02 -> 1.72 | Hurt |
| Transfer the CIFAR-10-searched cell to Freiburg | XNAS-Small · Freiburg test | top-1 test error (%): 10.7 -> 6.3 | Helped |
| Transfer the CIFAR-10-searched cell to CINIC-10 | XNAS-Small · CINIC-10 test | top-1 test error (%): 6.83 -> 6.0 | Helped |
| Use XNAS updates rather than gradient descent with softmax | architecture-weight optimization in the deterministic 2D toy problem · 2D toy problem | convergence to the minimum | Helped |
| Use XNAS updates rather than gradient descent with softmax | architecture-weight optimization in the imbalanced-initialization toy problem · 2D toy problem | progress to the minimum | Helped |
| Use XNAS updates rather than gradient descent with softmax | expert selection under noisy rewards · statistical expert-selection simulation | regret | Helped |
| Use XNAS updates rather than gradient descent with softmax | expert selection under noisy rewards · statistical expert-selection simulation | correct selection fraction | Helped |
Closest papers by meaning
- Adaptive Stochastic Natural Gradient Method for One-Shot Neural Architecture Search (ICML 2019 · similarity 0.91)
- Neural Architecture Search on ImageNet in Four GPU Hours: A Theoretically Inspired Perspective (ICLR 2021 · similarity 0.89)
- Hierarchical Representations for Efficient Architecture Search (ICLR 2018 · similarity 0.89)
- ISTA-NAS: Efficient and Consistent Neural Architecture Search by Sparse Coding (NeurIPS 2020 · similarity 0.88)
- MiLeNAS: Efficient Neural Architecture Search via Mixed-Level Reformulation (CVPR 2020 · similarity 0.88)
- EG-NAS: Neural Architecture Search with Fast Evolutionary Exploration (AAAI 2024 · 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.