IOTSC TALK SERIES: The Mathematical Evolution of AI Optimization

Dear Colleagues and Students,
The State Key Laboratory of Internet of Things for Smart City would like to invite you to join our “IOTSC Talk Series” on 27/04/2026 (Monday). We are pleased to invite Prof. Wenjing YAN from The Chinese University of Hong Kong as the speaker.
The Mathematical Evolution of AI Optimization
Speaker: Prof. Wenjing YAN
Date: 27/04/2026 (Monday)
Time: 15:00 – 16:00
Language: English
Venue: N21-6007
Abstract:
The rapid scaling of AI models has exposed the severe limitations and high computational costs of traditional, manually tuned optimization methods. This seminar explores the mathematical evolution of neural network training through two critical paradigm shifts. First, we examine True Parameter-Free Optimization, demonstrating how autonomous algorithms eliminate the need for heuristic hyperparameter tuning across diverse complex distributed systems by leveraging techniques such as normalization, momentum, and control variates. Second, we explore the Geometric Paradigm Shift required for trillion-parameter Large Language Models. By moving beyond standard optimizers that flatten 2D weight matrices into 1D arrays, we introduce advanced geometric methods (such as Muon and FISMO) that utilize matrix orthogonalization to respect the native multidimensional structure of neural networks. Ultimately, this talk outlines the future of AI training: replacing fragile scalar guesswork with robust, autonomous, and geometry-aware optimization.
Speaker’s Bio:
Dr. Wenjing Yan is currently a Research Assistant Professor in the Department of Information Engineering at The Chinese University of Hong Kong (CUHK). Prior to joining CUHK, she received her Ph.D. from the Department of Electronic and Computer Engineering at the Hong Kong University of Science and Technology (HKUST) in 2024. She obtained her M.S. degree from the University of Electronic Science and Technology of China (UESTC) in 2021 and B.Eng degree from Chongqing University in 2018. Dr. Yan’s research interests lie in distributed optimization, machine learning theory, and signal processing in communication and networking systems.
All are Welcome!
For enquiries: Tel: 8822 9976
Email: kennyfu@um.edu.mo
Best Regards,
State Key Laboratory of Internet of Things for Smart City