Publication Type

Conference Proceeding Article

Version

acceptedVersion

Publication Date

6-2020

Abstract

We investigate the compression of deep neural networks by quantizing their weights and activations into multiple binary bases, known as multi-bit networks (MBNs), which accelerate the inference and reduce the storage for the deployment on low-resource mobile and embedded platforms. We propose Adaptive Loss-aware Quantization (ALQ), a new MBN quantization pipeline that is able to achieve an average bitwidth below one-bit without notable loss in inference accuracy. Unlike previous MBN quantization solutions that train a quantizer by minimizing the error to reconstruct full precision weights, ALQ directly minimizes the quantizationinduced error on the loss function involving neither gradient approximation nor full precision maintenance. ALQ also exploits strategies including adaptive bitwidth, smooth bitwidth reduction, and iterative trained quantization to allow a smaller network size without loss in accuracy. Experiment results on popular image datasets show that ALQ outperforms state-of-the-art compressed networks in terms of both storage and accuracy.

Keywords

Quantization (signal), Optimization, Neural networks, Adaptive systems, Microprocessors, Training, Tensile stress

Discipline

Databases and Information Systems | Numerical Analysis and Scientific Computing

Research Areas

Software and Cyber-Physical Systems

Publication

2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition: June 16-18, Seattle, WA: Proceedings

First Page

7988

Last Page

7997

ISBN

9781728171692

Identifier

10.1109/CVPR42600.2020.00801

Publisher

IEEE Computer Society

City or Country

Los Alamitos, CA

Copyright Owner and License

Authors

Additional URL

https://doi.org/10.1109/CVPR42600.2020.00801

Download

Share

COinS