Neural Architecture Search
ProxylessNAS
Introduced by Cai et al. in ProxylessNAS: Direct Neural Architecture Search on Target Task and HardwareProxylessNAS directly learns neural network architectures on the target task and target hardware without any proxy task. Additional contributions include:
- Using a new path-level pruning perspective for neural architecture search, showing a close connection between NAS and model compression. Memory consumption is saved by one order of magnitude by using path-level binarization.
- Using a novel gradient-based approach (latency regularization loss) for handling hardware objectives (e.g. latency). Given different hardware platforms: CPU/GPU/Mobile, ProxylessNAS enables hardware-aware neural network specialization that’s exactly optimized for the target hardware.
Papers
| Paper | Code | Results | Date | Stars |
|---|
Tasks
| Task | Papers | Share |
|---|---|---|
| Image Classification | 4 | 22.22% |
| Federated Learning | 2 | 11.11% |
| Reinforcement Learning | 2 | 11.11% |
| Reinforcement Learning (RL) | 2 | 11.11% |
| Object Detection | 1 | 5.56% |
| Time Series Analysis | 1 | 5.56% |
| Meta-Learning | 1 | 5.56% |
| Diversity | 1 | 5.56% |
| Automatic Speech Recognition (ASR) | 1 | 5.56% |
Usage Over Time
Components
| Component | Type |
|
|---|---|---|
Adam
|
Stochastic Optimization | |
Cutout
|
Image Data Augmentation | |
DropPath
|
Regularization | |
REINFORCE
|
Policy Gradient Methods |
