App. Res. of Comp. Accepted Paper
Vol. 43, 2026 No. 6

Parameter tuning method for deep neural networks based on constraint solving and structural feature modeling

Xue Wen1a,1b,1c
Chen Hao2
Liang Yilei1a,1b,1c
Liu Baoying1a,1b,1c
Hu Ying3
Tang Zhanyong1a,1b,1c
1. Northwest University, a. School of Computer Science, b. Shaanxi Key Laboratory of Passive Internet of Things and Neural Computing, c. Xi'an Key Laboratory of Advanced Computing and Software Security, Xi'an 710127, China
2. China University of Labor Relations, Beijing 100048, China
3. Communications Technology Bureau of Xinhua News Agency, Beijing 100803, China
DOI: 10.19734/j.issn.1001-3695.2025.10.0431

Abstract

The rapid growth of deep neural networks makes the efficient execution of high-performance tensor programs essential for achieving low-latency inference. However, existing GPU-based auto-tuning approaches rely on large-scale search and suffer from invalid parameter configurations and insufficient feature representation in cost models, which leads to low tuning efficiency and poor performance prediction accuracy. This study aimed to improve the efficiency and accuracy of GPU tensor program auto-tuning by introducing a tuning method that integrates constraint solving and structural feature modeling. A set of parameter constraint rules for tensor programs was constructed using the Z3 solver to reduce invalid configurations and shrink the search space. In addition, to address the inability of traditional cost models to capture computational graph structure, a graph convolutional network was employed to extract structural features from the graph. These features were fused with traditional parameter features to build a high-quality training dataset, enhancing the model’s capability to predict the performance of different tensor program configurations. Experiments show that the proposed method accelerates inference on various deep neural networks by an average factor of 2.55 on NVIDIA GPUs compared with existing systems. Overall, the method significantly improves the efficiency of GPU tensor program tuning and the quality of performance prediction, providing a practical and effective approach for automated optimization in deep learning compilers.

Foundation Support

国家自然科学基金基金项目(62372373)
陕西省国际合作重点研发计划项目(2023-GHZD-04)
ResearchonAl-DrivenUnionOfficeDecision-MakingIntelligence(24XYJS018)

Publish Information

DOI: 10.19734/j.issn.1001-3695.2025.10.0431
Publish at: Application Research of Computers Accepted Paper, Vol. 43, 2026 No. 6

Publish History

[2026-02-26] Accepted Paper

Cite This Article

薛闻, 陈昊, 梁艺蕾, 等. 基于约束求解和结构特征建模的深度神经网络参数调优方法 [J]. 计算机应用研究, 2026, 43 (6). (2026-03-24). https://doi.org/10.19734/j.issn.1001-3695.2025.10.0431. (Xue Wen, Chen Hao, Liang Yilei, et al. Parameter tuning method for deep neural networks based on constraint solving and structural feature modeling [J]. Application Research of Computers, 2026, 43 (6). (2026-03-24). https://doi.org/10.19734/j.issn.1001-3695.2025.10.0431. )

About the Journal

  • Application Research of Computers Monthly Journal
  • Journal ID ISSN 1001-3695
    CN  51-1196/TP

Application Research of Computers, founded in 1984, is an academic journal of computing technology sponsored by Sichuan Institute of Computer Sciences under the Science and Technology Department of Sichuan Province.

Aiming at the urgently needed cutting-edge technology in this discipline, Application Research of Computers reflects the mainstream technology, hot technology and the latest development trend of computer application research at home and abroad in a timely manner. The main contents of the journal include high-level academic papers in this discipline, the latest scientific research results and major application results. The contents of the columns involve new theories of computer discipline, basic computer theory, algorithm theory research, algorithm design and analysis, blockchain technology, system software and software engineering technology, pattern recognition and artificial intelligence, architecture, advanced computing, parallel processing, database technology, computer network and communication technology, information security technology, computer image graphics and its latest hot application technology.

Application Research of Computers has many high-level readers and authors, and its readers are mainly senior and middle-level researchers and engineers engaged in the field of computer science, as well as teachers and students majoring in computer science and related majors in colleges and universities. Over the years, the total citation frequency and Web download rate of Application Research of Computers have been ranked among the top of similar academic journals in this discipline, and the academic papers published are highly popular among the readers for their novelty, academics, foresight, orientation and practicality.

Indexed & Evaluation

  • The Second National Periodical Award 100 Key Journals
  • Double Effect Journal of China Journal Formation
  • the Core Journal of China (Peking University 2023 Edition)
  • the Core Journal for Science
  • Chinese Science Citation Database (CSCD) Source Journals
  • RCCSE Chinese Core Academic Journals
  • Journal of China Computer Federation
  • 2020-2022 The World Journal Clout Index (WJCI) Report of Scientific and Technological Periodicals
  • Full-text Source Journal of China Science and Technology Periodicals Database
  • Source Journal of China Academic Journals Comprehensive Evaluation Database
  • Source Journals of China Academic Journals (CD-ROM Version), China Journal Network
  • 2017-2019 China Outstanding Academic Journals with International Influence (Natural Science and Engineering Technology)
  • Source Journal of Top Academic Papers (F5000) Program of China's Excellent Science and Technology Journals
  • Source Journal of China Engineering Technology Electronic Information Network and Electronic Technology Literature Database
  • Source Journal of British Science Digest (INSPEC)
  • Japan Science and Technology Agency (JST) Source Journal
  • Russian Journal of Abstracts (AJ, VINITI) Source Journals
  • Full-text Journal of EBSCO, USA
  • Cambridge Scientific Abstracts (Natural Sciences) (CSA(NS)) core journals
  • Poland Copernicus Index (IC)
  • Ulrichsweb (USA)