Research map / Architectures and efficiency / Neural architecture search
ChamNet: Towards Efficient Network Design Through Platform-Aware Model Adaptation
Xiaoliang Dai, Peizhao Zhang, Bichen Wu, Hongxu Yin, Fei Sun, Yanghan Wang, Marat Dukhan, Yunqing Hu, Yiming Wu, Yangqing Jia, Peter Vajda, Matt Uyttendaele and 1 more
Check your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 |
|---|---|---|---|
| adapt MobileNetV2 building blocks to fit the target platform and latency budget | ChamNet-Mobile · ImageNet, Snapdragon 835 CPU, 4.0 ms | top-1 accuracy: 8.5% absolute accuracy gain | Improved |
| adapt MobileNetV2 building blocks to fit the target platform and latency budget | ChamNet-Mobile · ImageNet, Snapdragon 835 CPU, 20 ms | top-1 accuracy: 73.8% | Improved |
| adapt MobileNetV2 building blocks to fit the target platform and latency budget | ChamNet-Mobile · ImageNet, Hexagon v62 DSP, 20 ms | top-1 accuracy: 75.3% | Improved |
| adapt MobileNetV2 building blocks for energy-constrained inference | ChamNet-Mobile · ImageNet, Snapdragon 835 CPU, 14mJ per run | top-1 accuracy: 58.8% -> 60.0% | Improved |
| adapt MobileNetV2 building blocks for energy-constrained inference | ChamNet-Mobile · ImageNet, Snapdragon 835 CPU | energy per run: 19mJ -> 14mJ | Improved |
| redistribute computation from early stages toward later stages | ChamNet-Mobile · ImageNet, Snapdragon 835 CPU | top-1 accuracy: 69.8% -> 71.9% | Improved |
| use direct platform latency measurements rather than FLOPs as the optimization target | Chameleon · mobile CPU and DSP | latency | Improved |
| select accuracy-predictor training samples using Bayesian optimization with exploration and exploitation | Chameleon accuracy predictor · MobileNetV2 accuracy prediction | sample efficiency | Improved |
| use an operator latency lookup table for candidate latency estimation | Chameleon latency predictor · target-device network latency estimation | estimation time: less than one CPU second | Improved |
Closest papers by meaning
- Chameleon: Adaptive Code Optimization for Expedited Deep Neural Network Compilation (ICLR 2020 · similarity 0.89)
- Resource Constrained Neural Network Architecture Search: Will a Submodularity Assumption Help? (ICCV 2019 · similarity 0.87)
- Rapid Model Architecture Adaption for Meta-Learning (NeurIPS 2022 · similarity 0.86)
- Efficient Architecture Search by Network Transformation (AAAI 2018 · similarity 0.86)
- TAPAS: Train-Less Accuracy Predictor for Architecture Search (AAAI 2019 · similarity 0.86)
- EG-NAS: Neural Architecture Search with Fast Evolutionary Exploration (AAAI 2024 · similarity 0.84)
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.
