Research map / Architectures and efficiency / Neural architecture search
iDARTS: Differentiable Architecture Search with Stochastic Implicit Gradients
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 |
|---|---|---|---|
| Replace one-step unrolled DARTS hypergradients with stochastic implicit gradients | DARTS architecture search · NAS-Bench-201 CIFAR-10 | test accuracy: 54.30±0.00 -> 93.58±0.32 | Helped |
| Replace one-step unrolled DARTS hypergradients with stochastic implicit gradients | DARTS architecture search · NAS-Bench-201 CIFAR-100 | test accuracy: 15.61±0.00 -> 70.83±0.48 | Helped |
| Replace one-step unrolled DARTS hypergradients with stochastic implicit gradients | DARTS architecture search · NAS-Bench-201 ImageNet-16-120 | test accuracy: 16.32±0.00 -> 40.89±0.68 | Helped |
| Use iDARTS instead of DARTS | DARTS search space · CIFAR-10 | test error: 2.76±0.09 -> 2.37±0.03 | Helped |
| Use iDARTS instead of DARTS | DARTS search space · CIFAR-100 | test error: 17.54 -> 16.02 | Helped |
| Use iDARTS instead of DARTS | DARTS search space · ImageNet | top1/top5 test error: 26.9 / 8.7 -> 24.3 / 7.3 | Mixed |
| Increase inner optimization steps from T=1 to T=4 (ablation) | iDARTS · NAS-Bench-201 CIFAR-10 | test accuracy: 90.36±1.79 -> 92.44±1.14 | Helped |
| Increase Neumann approximation terms from K=0 to K=2 (ablation) | iDARTS with T=4 · NAS-Bench-201 CIFAR-10 | test accuracy: 90.36±1.79 -> 93.58±0.32 | Helped |
| Increase Neumann approximation terms from K=2 to K=3 (ablation) | iDARTS with T=4 · NAS-Bench-201 CIFAR-10 | test accuracy: 93.58±0.32 -> 92.29±0.26 | Hurt |
| Increase Neumann approximation terms from K=0 to K=2 (ablation) | iDARTS with T=4 · NAS-Bench-201 CIFAR-100 | test accuracy: 64.43±2.47 -> 70.83±0.48 | Helped |
| Increase Neumann approximation terms from K=0 to K=2 (ablation) | iDARTS with T=4 · NAS-Bench-201 ImageNet-16-120 | test accuracy: 32.42±1.54 -> 40.89±0.68 | Helped |
| Increase inner optimization learning rate γ (ablation) | iDARTS · NAS-Bench-201 | performance | Hurt |
Closest papers by meaning
- BaLeNAS: Differentiable Architecture Search via the Bayesian Learning Rule (CVPR 2022 · similarity 0.94)
- Rethinking Architecture Selection in Differentiable NAS (ICLR 2021 · similarity 0.93)
- EG-NAS: Neural Architecture Search with Fast Evolutionary Exploration (AAAI 2024 · similarity 0.93)
- Differentiable Architecture Search With Random Features (CVPR 2023 · similarity 0.92)
- $\Lambda$-DARTS: Mitigating Performance Collapse by Harmonizing Operation Selection among Cells (ICLR 2023 · similarity 0.92)
- IS-DARTS: Stabilizing DARTS through Precise Measurement on Candidate Importance (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.