Research map / Architectures and efficiency / CNN design
Designing Neural Network Architectures using Reinforcement Learning
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 MetaQNN architecture search instead of the strongest reported benchmark baseline | MetaQNN · CIFAR-10 | error rate (%): 4.62 -> 6.92 | Hurt |
| Use MetaQNN architecture search instead of the strongest reported benchmark baseline | MetaQNN · SVHN | error rate (%): 1.69 -> 2.06 | Hurt |
| Use MetaQNN architecture search instead of the strongest reported benchmark baseline | MetaQNN · MNIST | error rate (%): 0.31 -> 0.28% | Mixed |
| Use MetaQNN architecture search instead of the strongest reported benchmark baseline | MetaQNN · CIFAR-100 | error rate (%): 22.71 -> 27.14$^{\boldsymbol{*}}$ | Hurt |
| Decrease ε from 1.0 to 0.1 so the agent increasingly exploits learned Q-values | MetaQNN on SVHN · SVHN | mean accuracy: 52.25% -> 88.02% | Helped |
| Ensemble all ten selected models rather than use the reported five-model ensemble | MetaQNN on MNIST · MNIST | error rate (%): 0.32 -> 0.28% | Helped |
| Ensemble the top five models rather than use the top model | MetaQNN on CIFAR-10 · CIFAR-10 | error rate (%): 6.92 -> 7.32 | Hurt |
Closest papers by meaning
- Efficient Architecture Search by Network Transformation (AAAI 2018 · similarity 0.88)
- Model-Agnostic Meta-Learning for Fast Adaptation of Deep Networks (ICML 2017 · similarity 0.87)
- Neural Architecture Optimization (NeurIPS 2018 · similarity 0.87)
- Neural Architecture Search with Reinforcement Learning (ICLR 2017 · similarity 0.86)
- Path-Level Network Transformation for Efficient Architecture Search (ICML 2018 · similarity 0.86)
- Jointly Training and Pruning CNNs via Learnable Agent Guidance and Alignment (CVPR 2024 · similarity 0.85)
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.