Research map / Probabilistic and scientific modeling / Variational and generative sampling
Differentiable Neural Architecture Search in Equivalent Space with Exploration Enhancement
Miao Zhang, Huiqi Li, Shirui Pan, Xiaojun Chang, Zongyuan Ge, Steven Su
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 continuous relaxation (GDAS-A) with injective graph-autoencoder latent space, no exploration or complementation (ablation) | E2NAS · NAS-Bench-201 CIFAR-10 | test accuracy (%): 89.73 -> 92.75 | Improved |
| same replacement on ImageNet-16-120 (ablation) | E2NAS · NAS-Bench-201 ImageNet-16-120 | test accuracy (%): 35.34 -> 43.19 | Improved |
| add sigmoid-scheduled exploration enhancement Sig(1) to E2NAS (ablation) | E2NAS · NAS-Bench-201 ImageNet-16-120 | test accuracy (%): 43.19 -> 46.10 | Improved |
| add sigmoid-scheduled exploration Sig(2) to E2NAS (ablation) | E2NAS · NAS-Bench-201 CIFAR-10 | best accuracy over search (%): 93.79 -> 94.29 | Improved |
| add static exploration gamma=0.2 (ablation) | E2NAS · NAS-Bench-201 CIFAR-10 | test accuracy (%): 92.75 -> 82.53 | Hurt |
| add static exploration gamma=0.5 (ablation) | E2NAS · NAS-Bench-201 ImageNet-16-120 | test accuracy (%): 43.19 -> 44.43 | Improved |
| add architecture complementation loss Lc (medium epsilon 0.2 or 0.5) (ablation) | E2NAS, RandomNAS, GDAS · E2NAS, RandomNAS, GDAS | NAS-Bench-201 accuracy | Improved |
| add Lc to GDAS (GDAS-AC) to relieve catastrophic forgetting (ablation) | GDAS · GDAS | DARTS search space validation accuracy during supernet training | Improved |
Closest papers by meaning
- Rethinking Architecture Selection in Differentiable NAS (ICLR 2021 · similarity 0.90)
- BaLeNAS: Differentiable Architecture Search via the Bayesian Learning Rule (CVPR 2022 · similarity 0.90)
- Rethinking Bi-Level Optimization in Neural Architecture Search: A Gibbs Sampling Perspective (AAAI 2021 · similarity 0.90)
- Overcoming Multi-Model Forgetting in One-Shot NAS With Diversity Maximization (CVPR 2020 · similarity 0.89)
- Efficient Neural Architecture Search via Proximal Iterations (AAAI 2020 · 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.
