Network Space Search for Pareto-Efficient Spaces
Abstract
Network spaces have been known as a critical factor in both handcrafted network designs or defining search spaces for Neural Architecture Search (NAS). However, an effective space involves tremendous prior knowledge and/or manual effort, and additional constraints are required to discover efficiency-aware architectures. In this paper, we define a new problem, Network Space Search (NSS), as searching for favorable network spaces instead of a single architecture. We propose an NSS method to directly search for efficient-aware network spaces automatically, reducing the manual effort and immense cost in discovering satisfactory ones. The resultant network spaces, named Elite Spaces, are discovered from Expanded Search Space with minimal human expertise imposed. The Pareto-efficient Elite Spaces are aligned with the Pareto front under various complexity constraints and can be further served as NAS search spaces, benefiting differentiable NAS approaches (e.g. In CIFAR-100, an averagely 2.3% lower error rate and 3.7% closer to target constraint than the baseline with around 90% fewer samples required to find satisfactory networks). Moreover, our NSS approach is capable of searching for superior spaces in future unexplored spaces, revealing great potential in searching for network spaces automatically.
Video
2-min talk video:
Resources
 |
|---|
Other Links:
- CVF Open Access
- IEEE Xplore (to be updated)
- CVPR workshop on Efficient Deep Learning for Computer Vision (ECV) [ CVPRW'21 ]
- MediaTek Blog
Citation
Min-Fong Hong, Hao-Yun Chen, Min-Hung Chen, Yu-Syuan Xu, Hsien-Kai Kuo, Yi-Min Tsai, Hung-Jen Chen, and Kevin Jou, “Network Space Search for Pareto-Efficient Spaces “, CVPR Workshop on Efficient Deep Learning for Computer Vision (CVPRW), 2021 [Oral].
BibTex
@inproceedings{hong2021network,
title={Network Space Search for Pareto-Efficient Spaces},
author={Hong, Min-Fong and Chen, Hao-Yun and Chen, Min-Hung and Xu, Yu-Syuan and Kuo, Hsien-Kai and Tsai, Yi-Min and Chen, Hung-Jen and Jou, Kevin},
booktitle={IEEE Conference on Computer Vision and Pattern Recognition (CVPR) Workshops (Efficient Deep Learning for Computer Vision)},
year={2021},
url={https://arxiv.org/abs/2104.11014}
}
Members
MediaTek Inc.
 |  |  |  |  |  |  |  |
|---|
