Binarized Neural Architecture Search for Efficient Object Recognition
- PDF / 1,642,507 Bytes
- 16 Pages / 595.276 x 790.866 pts Page_size
- 46 Downloads / 202 Views
Binarized Neural Architecture Search for Efficient Object Recognition Hanlin Chen1 · Li’an Zhuo1 · Baochang Zhang1,2 David Doermann5 · Guodong Guo6,7
· Xiawu Zheng3 · Jianzhuang Liu4 · Rongrong Ji3 ·
Received: 19 December 2019 / Accepted: 28 August 2020 © Springer Science+Business Media, LLC, part of Springer Nature 2020
Abstract Traditional neural architecture search (NAS) has a significant impact in computer vision by automatically designing network architectures for various tasks. In this paper, binarized neural architecture search (BNAS), with a search space of binarized convolutions, is introduced to produce extremely compressed models to reduce huge computational cost on embedded devices for edge computing. The BNAS calculation is more challenging than NAS due to the learning inefficiency caused by optimization requirements and the huge architecture space, and the performance loss when handling the wild data in various computing applications. To address these issues, we introduce operation space reduction and channel sampling into BNAS to significantly reduce the cost of searching. This is accomplished through a performance-based strategy that is robust to wild data, which is further used to abandon less potential operations. Furthermore, we introduce the upper confidence bound to solve 1-bit BNAS. Two optimization methods for binarized neural networks are used to validate the effectiveness of our BNAS. Extensive experiments demonstrate that the proposed BNAS achieves a comparable performance to NAS on both CIFAR and ImageNet databases. An accuracy of 96.53% vs. 97.22% is achieved on the CIFAR-10 dataset, but with a significantly compressed model, and a 40% faster search than the state-of-the-art PC-DARTS. On the wild face recognition task, our binarized models achieve a performance similar to their corresponding full-precision models. Keywords Neural architecture search (NAS) · Binarized network · Object recognition · Edge computing
1 Introduction Efficient computing has become one of the hottest topics both in academy and industry. It will be vital for the 5G networks by providing hardware-friendly and efficient solutions for practical and wild applications (Mao et al. 2017). Edge
computing is about computing resources that are closer to the end user. This makes applications faster and users friendly (Chen and Ran 2019). It enables mobile or embedded devices to provide real-time intelligent analysis of big data, which can reduce the pressure on the cloud computing center and improve the availability (Han et al. 2019). However, edge
Communicated by Cha Zhang.
B
Baochang Zhang [email protected]
Guodong Guo [email protected] 1
Beihang University, Beijing, China
Hanlin Chen [email protected]
2
Shenzhen Academy of Aerospace Technology, Shenzhen 100083, China
Li’an Zhuo [email protected]
3
Xiamen University, Xiamen, Fujian, China
4
Shenzhen Institutes of Advanced Technology, Shenzhen, China
5
University at Buffalo, Buffalo, NY, USA
6
Institute of Deep Learning, Baidu Research, Bei
Data Loading...
