Research map / Architectures and efficiency / Neural architecture search
On Redundancy and Diversity in Cell-based Neural Architecture Search
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 |
|---|---|---|---|
| Constrain random cell sampling to include residual skip links and use separable convolutions for other operations | DARTS cell search space · CIFAR10 test | Top-1 test error (%): 5.24%; only <1% of the training set perform better than the mean predicted test error | Helped |
| Constrain random cell sampling to include residual skip links and use separable convolutions for other operations | DARTS cell search space · ImageNet mobile test | Top-1 test error (%): 23.9 | Helped |
| Edit existing cells to comply with PrimSkip patterns (ablation) | DARTS-cell NAS architectures · CIFAR10 test | Top-1 test error (%): -0.08; +0.03; +0.04; -0.13; -0.10; -0.15; +0.02; -0.10; -0.44 | Mixed |
| Set the reduce cell identical to the normal cell (ablation) | DARTS architectures · CIFAR10 | test performance | No clear effect |
| Set the normal cell identical to the reduce cell (ablation) | DARTS architectures · CIFAR10 | test performance | Hurt |
| Replace reduce-cell operations with skip connections (ablation) | DARTS architectures · CIFAR10 | test performance | Hurt |
| Replace normal-cell operations with skip connections (ablation) | DARTS architectures · CIFAR10 | test performance | Hurt |
| Remove low-importance operations (ablation) | Top-performing DARTS architectures · CIFAR10 | accuracy: 6 low-OI operations | No clear effect |
| Remove high-importance operations (ablation) | Top-performing DARTS architectures · CIFAR10 | accuracy: 2 high-OI operations | Hurt |
| Use the Skip constraint alone (ablation) | NAS-Bench-201 · CIFAR10, CIFAR100, ImageNet16-120 | test performance | No clear effect |
| Use both Prim and Skip constraints (ablation) | NAS-Bench-201 · CIFAR10, CIFAR100, ImageNet16-120 | test performance: fewer than 50 samples | Helped |
| Add pooling operations to PrimSkip sampling (ablation) | PrimSkip search space · CIFAR10 | accuracy-cost Pareto front: 1,000 architectures | No clear effect |
Closest papers by meaning
- Understanding Architectures Learnt by Cell-based Neural Architecture Search (ICLR 2020 · similarity 0.93)
- einspace: Searching for Neural Architectures from Fundamental Operations (NeurIPS 2024 · similarity 0.90)
- Towards Neural Architecture Search through Hierarchical Generative Modeling (ICML 2024 · similarity 0.89)
- Efficient Few-Shot Neural Architecture Search by Counting the Number of Nonlinear Functions (AAAI 2025 · similarity 0.87)
- EG-NAS: Neural Architecture Search with Fast Evolutionary Exploration (AAAI 2024 · similarity 0.87)
- Insights from the Use of Previously Unseen Neural Architecture Search Datasets (CVPR 2024 · similarity 0.87)
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.