Research map / Architectures and efficiency / Neural architecture search
Neural Architecture Search Using Deep Neural Networks and Monte Carlo Tree 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 |
|---|---|---|---|
| Use AlphaX search with cutout and 32 filters | CIFAR-10 NAS · CIFAR-10 | test error: 2.50 -> 2.54 | Helped |
| Use AlphaX without cutout at 32 filters | CIFAR-10 NAS · CIFAR-10 | test error: 3.18 -> 3.04 | Helped |
| Increase the filter count from 32 to 128 while using cutout (ablation) | AlphaX · CIFAR-10 | test error: $2.54_{\pm0.06}$ -> $2.16_{\pm0.04}$ | Helped |
| Use AlphaX-1 rather than RENASNet | ImageNet mobile setting · ImageNet | top-1 error: 24.3 -> 24.5 | Hurt |
| Use MCTS-guided AlphaX search instead of Random Search | NASBench-101 search · NASBench-101 (nodes ≤ 6) | sample efficiency: 3x | Helped |
| Use MCTS-guided AlphaX search instead of Regularized Evolution | NASBench-101 search · NASBench-101 (nodes ≤ 6) | sample efficiency: 2.8x | Helped |
| Use an MLP rather than an RNN as the meta-DNN predictor (ablation) | AlphaX meta-DNN · NASBench-101 | prediction-accuracy correlation: corr=0.743 -> corr=0.784 | Helped |
| Use a multi-stage MLP with specialists for accuracy ranges (ablation) | AlphaX meta-DNN · NASBench-101 | prediction-accuracy correlation: 1.2% | Helped |
| Transfer parent-network weights to overlapping child layers | AlphaX network evaluation · Child network training | epochs to reach same accuracy: Transfer 17\rightarrow70 epochs v.s. random initialization | Helped |
| Reduce transfer-learning pretraining to 9 epochs (ablation) | AlphaX network evaluation · Child network training | accuracy: 9 epochs | Hurt |
| Replace MobileNet-v1 with AlphaX-1 as the SSD backbone | SSD object detection · MSCOCO mini-val, 300×300 | mAP: 20.1% -> 23.7% | Helped |
Closest papers by meaning
- Prioritized Architecture Sampling With Monto-Carlo Tree Search (CVPR 2021 · similarity 0.92)
- Neural Architecture Search on ImageNet in Four GPU Hours: A Theoretically Inspired Perspective (ICLR 2021 · similarity 0.90)
- ISTA-NAS: Efficient and Consistent Neural Architecture Search by Sparse Coding (NeurIPS 2020 · similarity 0.89)
- MOTE-NAS: Multi-Objective Training-based Estimate for Efficient Neural Architecture Search (NeurIPS 2024 · similarity 0.88)
- Towards Neural Architecture Search through Hierarchical Generative Modeling (ICML 2024 · similarity 0.88)
- Efficient Few-Shot Neural Architecture Search by Counting the Number of Nonlinear Functions (AAAI 2025 · 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.