Title:
GRAPH NEURAL NETWORK COMPRESSION METHOD AND APPARATUS, AND ELECTRONIC DEVICE AND STORAGE MEDIUM
Document Type and Number:
WIPO Patent Application WO/2024/087512
Kind Code:
A1
Abstract:
The present application relates to the field of neural networks. Provided are a graph neural network compression method and apparatus, and an electronic device and a storage medium. The method comprises: acquiring a trained graph neural network and graph data used during the training of the graph neural network; determining a degree distribution range corresponding to all graph vertexes in the graph data, and dividing the degree distribution range into a plurality of degree sections; under the constraint of a preset resource limiting condition, using reinforcement learning and a hardware accelerator to determine an optimal section quantization bit width corresponding to each degree section and an optimal network quantization bit width corresponding to the graph neural network; and using the optimal section quantization bit width to perform quantization compression on vertex features of the graph vertexes of corresponding degrees in the graph data, and using the optimal network quantization bit width to perform quantization compression on the graph neural network, so as to obtain optimal quantization graph data and an optimal quantization graph neural network. Therefore, optimal quantization bit widths are determined for a graph neural network and graph vertex features by using reinforcement learning, so as to ensure that a quantization graph neural network has high precision and a relatively low resource consumption rate.
Inventors:
HU KEKUN (CN)
DONG GANG (CN)
ZHAO YAQIAN (CN)
LI RENGANG (CN)
DONG GANG (CN)
ZHAO YAQIAN (CN)
LI RENGANG (CN)
Application Number:
PCT/CN2023/085970
Publication Date:
May 02, 2024
Filing Date:
April 03, 2023
Export Citation:
Assignee:
INSPUR ELECTRONIC INFORMATION INDUSTRY CO LTD (CN)
International Classes:
G06F16/174; G06F9/50; G06N3/04; G06N3/08
Foreign References:
| CN115357554A | 2022-11-18 | |||
| CN111563589A | 2020-08-21 | |||
| CN110852439A | 2020-02-28 | |||
| CN113570037A | 2021-10-29 | |||
| CN113762489A | 2021-12-07 | |||
| CN113902108A | 2022-01-07 | |||
| US20190340492A1 | 2019-11-07 | |||
| US20220092391A1 | 2022-03-24 |
Other References:
YIN WEN-FENG, LIANG LING-YAN , PENG HUI-MIN , CAO QI-CHUN , ZHAO JIAN , DONG GANG , ZHAO YA-QIAN , ZHAO KUN: "Research Progress on Convolutional Neural Network Compression and Acceleration Technology", COMPUTER SYSTEMS AND APPLICATIONS, ZHONGGUO KEXUEYUAN RUANJIAN YANJIUSUO, CN, vol. 29, no. 9, 15 September 2020 (2020-09-15), CN , pages 16 - 25, XP093028237, ISSN: 1003-3254, DOI: 10.15888/j.cnki.csa.007632
Attorney, Agent or Firm:
BEIJING RUN ZEHENG INTELLECTUAL PROPERTY LAW FIRM (CN)
Download PDF: