🔬 Research Interests
Gauss–Seidel vs. Jacobi Iteration
I am currently interested in one central question: the contrasting convergence behavior of Gauss–Seidel (GS) and Jacobi (JC) iterations. This question has accompanied me throughout my research career. It originated from updating matrices A and B in LoRA, where we observed that GS converges significantly faster than JC during the first half of training, yet ultimately achieves worse convergence.
This puzzling phenomenon has also been consistently observed in deep learning, federated learning, and clustering. We have therefore continued to investigate the underlying reason.
Recently, we finally obtained some theoretical insights. If you are interested in this question, please read our paper below. We will continue to explore it in greater depth.
📄 Read the Paper📚 Publications
🎓 Education
Ph.D. - Computer Science and Technology
Wuhan University | 2023.09-2026.06
Advised by Prof. Sheng Wang and Prof. Zhiyong Peng
B.S. - Electronic Information Engineering
Dalian University of Technology | 2014.09-2018.06
🏆 Honors & Awards
National Scholarship
Wuhan University | 2023
DiDi Scholarship Second Prize
Wuhan University | 2024
DiDi Scholarship Third Prize
Wuhan University | 2025
Huawei Scholarship
Wuhan University | 2025
🐱 Misc.
🏠 My Feline Family
These are my beloved cats who accompany me through long research hours. They are not just pets, but faithful companions who bring joy and inspiration to my academic journey.
📖 Academic Service
Journal Reviewer
JMLR, IEEE TKDE, ACM TKDD