Research map / Architectures and efficiency / Neural architecture search
Learning Transferable Architectures for Scalable Image Recognition
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 |
|---|---|---|---|
| Transfer the searched convolutional cells from CIFAR-10 to ImageNet and scale the network | NASNet-A · ImageNet 2012 ILSVRC | Top 1 Acc. (%): 82.7 | Helped |
| Add cutout data augmentation (ablation) | NASNet-A (7 @ 2304) · CIFAR-10 test | error rate (%): 2.97 -> 2.40 | Helped |
| Use ScheduledDropPath with linearly increasing path-drop probability instead of fixed-probability DropPath (ablation) | NASNet · CIFAR-10 and ImageNet | classification performance | Helped |
| Use fixed-probability DropPath instead of no path dropping (ablation) | NASNet · CIFAR-10 and ImageNet | classification performance | No clear effect |
| Manually insert residual connections between convolutional cells (ablation) | NASNet · ImageNet classification | Top 1 Acc. (%) | No clear effect |
| Search architectures with reinforcement learning rather than random search | NASNet search space · CIFAR-10 validation | validation performance: over 1% | Helped |
| Apply naive dropout across convolutional filters (ablation) | NASNet · CIFAR-10 and ImageNet | classification performance | Hurt |
| Apply separable convolutions twice to a hidden state rather than once (ablation) | NASNet cells · image classification | classification performance | Helped |
| Use ELU instead of ReLU (ablation) | NASNet search-space convolutions · image classification | classification performance | No clear effect |
Closest papers by meaning
- TransNAS-Bench-101: Improving Transferability and Generalizability of Cross-Task Neural Architecture Search (CVPR 2021 · similarity 0.89)
- NAS-Bench-360: Benchmarking Neural Architecture Search on Diverse Tasks (NeurIPS 2022 · similarity 0.89)
- Learning Versatile Neural Architectures by Propagating Network Codes (ICLR 2022 · similarity 0.89)
- Rapid Model Architecture Adaption for Meta-Learning (NeurIPS 2022 · similarity 0.89)
- Training-free Neural Architecture Search through Variance of Knowledge of Deep Network Weights (CVPR 2025 · similarity 0.88)
- L-SWAG: Layer-Sample Wise Activation with Gradients Information for Zero-Shot NAS on Vision Transformers (CVPR 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.