Publication Type
Conference Proceeding Article
Version
publishedVersion
Publication Date
9-2024
Abstract
Quantization plays an important role in deploying neural networks on embedded, real-time systems with limited computing and storage resources (e.g., edge devices). It significantly reduces the model storage cost and improves inference efficiency by using fewer bits to represent the parameters. However, it was recently shown that critical properties may be broken after quantization, such as robustness and backdoor-freeness. In this work, we introduce the first method for synthesizing quantization strategies that verifiably maintain desired properties after quantization, leveraging a key insight that quantization leads to a data distribution shift in each layer. We propose to compute the preimage for each layer based on which the preceding layer is quantized, ensuring that the quantized reachable region of the preceding layer remains within the preimage. To tackle the challenge of computing the exact preimage, we propose an MILP-based method to compute its under-approximation. We implement our method into a tool Quadapter and demonstrate its effectiveness and efficiency by providing certified quantization that successfully preserves model robustness and backdoor-freeness.
Discipline
Databases and Information Systems | Numerical Analysis and Scientific Computing
Research Areas
Intelligent Systems and Optimization
Publication
Formal Methods: 26th International Symposium FM 2024, Milan, September 9-13: Proceedings
Volume
14933
First Page
343
Last Page
362
ISBN
9783031711619
Identifier
10.1007/978-3-031-71162-6_18
Publisher
Springer
City or Country
Cham
Citation
ZHANG, Yedi; CHEN, Guangke; SUN, Jun; and SUN, Jun.
Certified quantization strategy synthesis for neural networks. (2024). Formal Methods: 26th International Symposium FM 2024, Milan, September 9-13: Proceedings. 14933, 343-362.
Available at: https://ink.library.smu.edu.sg/sis_research/9325
Copyright Owner and License
Authors
Creative Commons License
This work is licensed under a Creative Commons Attribution-NonCommercial-No Derivative Works 4.0 International License.
Additional URL
https://doi.org/10.1007/978-3-031-71162-6_18
Included in
Databases and Information Systems Commons, Numerical Analysis and Scientific Computing Commons