Research map / Architectures and efficiency / Neural architecture search
Understanding and Enhancing Differentiable Architecture Search from Information Bottleneck Perspective
Haidong Kang, Lianbo Ma, Pengjun Chen, Qiang He, Bo Yi
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 |
|---|---|---|---|
| add Batch Entropy-decay Regularization (BER) to DARTS | DARTS(1st) · NAS-Bench-201 CIFAR-10 | test accuracy (%): 54.30 -> 94.36 | Improved |
| add BER to DARTS | DARTS(1st) · NAS-Bench-201 CIFAR-100 | test accuracy (%): 15.61 -> 73.51 | Improved |
| add BER to DARTS | DARTS(1st) · NAS-Bench-201 ImageNet16-120 | test accuracy (%): 16.32 -> 46.34 | Improved |
| add BER versus beta-DARTS regularisation | beta-DARTS · NAS-Bench-201 | search cost (GPU-hours): 3.20 -> 0.16 | Improved |
| add BER versus beta-DARTS regularisation | beta-DARTS · NAS-Bench-201 CIFAR-10 | test accuracy (%): 94.36 -> 94.37 | No clear effect |
| use lambda 0.75 with 5 epochs instead of lambda 0.50 with 15 epochs (ablation) | r-DARTS · NAS-Bench-201 CIFAR-100 | test accuracy (%): 73.51 -> 73.09 | Mixed |
| search on CIFAR-100 instead of CIFAR-10 (ablation) | r-DARTS (λ=0.75) · NAS-Bench-201 ImageNet16-120 | test accuracy (%): 46.33 -> 47.31 | Improved |
| search on CIFAR-100 instead of CIFAR-10 (ablation) | r-DARTS (λ=0.75) · NAS-Bench-201 CIFAR-10 | test accuracy (%): 94.37 -> 93.99 | Hurt |
| add BER to DARTS | DARTS · ImageNet1k (DARTS search space) | top-1 accuracy (%): 73.3 -> 76.9 | Improved |
| add BER to DARTS | DARTS · DARTS search space memory | GPU memory: 8554 -> 8554 | No clear effect |
| add BER to DARTS-based search on MobileNet-like space versus ZiCo | ZiCo · ImageNet1k (MobileNet-like) | top-1 accuracy (%): 78.1 -> 78.9 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 4 of 4 on most shared results.
- β-DARTS — DARTS-: Robustly Stepping out of Performance Collapse Without Indicators (ICLR 2021) · better on 8 of 9 shared results
- ParZC — ParZC: Parametric Zero-Cost Proxies for Efficient NAS (AAAI 2025) · better on 2 of 2 shared results
- ZiCo — ZiCo: Zero-shot NAS via inverse Coefficient of Variation on Gradients (ICLR 2023) · better on 1 of 1 shared result
- DrNAS — DrNAS: Dirichlet Neural Architecture Search (ICLR 2021) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Stabilizing Differentiable Architecture Search via Perturbation-based Regularization (ICML 2020 · similarity 0.89)
- Differentiable Architecture Search With Random Features (CVPR 2023 · similarity 0.88)
- DARTS-: Robustly Stepping out of Performance Collapse Without Indicators (ICLR 2021 · similarity 0.88)
- $\beta $-DARTS++: Bi-Level Regularization for Proxy-Robust Differentiable Architecture Search (TPAMI 2025 · similarity 0.88)
- Understanding and Robustifying Differentiable Architecture Search (ICLR 2020 · similarity 0.88)
- BaLeNAS: Differentiable Architecture Search via the Bayesian Learning Rule (CVPR 2022 · 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.
