Academic Report of School of Mathematical Sciences [2026] No. 034

(Series Report for High-Level University Construction No. 1293)

Title：Mathematical Foundations of Next-Generation Machine Learning

Speaker：Deyu Meng, Professor (Xi'an Jiao Tong University)

Time：14:00-15:00, Apr. 27, 2026

Location：Tencent Meeting ID 919-597-901

Abstract：Machine learning methods and technologies, represented by deep learning and large-scale models, are currently at the forefront of scientific research. Recently, the performance of the Deepseek product has drawn widespread attention to the fields of machine learning and artificial intelligence. However, while technological capabilities have advanced rapidly, progress in fundamental theoretical research on machine learning has lagged far behind. Many empirical findings lack theoretical grounding, placing the scientific foundation of the discipline at risk. Rebuilding the theoretical framework of machine learning has thus become a major frontier issue in contemporary science and technology. In response to this challenge, this report will analyze the theoretical implications of typical experimental phenomena in deep learning—specifically task generalization, the emergence of intelligence, and the robustness-accuracy trade-off. It will introduce novel theoretical frameworks in learning, statistics, and physics that may shape the future of machine learning theory, as well as the development opportunities these frameworks could bring to addressing technical challenges in the fields of communications, life sciences, and medical technology, such as dynamic adaptation, the emergence of biological intelligence, and reliable intelligent diagnosis and treatment.

Speaker Profile：Deyu Meng is a professor at the School of Mathematics and Statistics, Xi'an Jiaotong University. He has long been dedicated to research on the fundamental theories and algorithms of machine learning. Over the past five years, he has published more than 100 papers in journals and conferences related to machine learning and has been selected as a Highly Cited Researcher by Clarivate and Elsevier for multiple years. He was selected for the Central Organization Department's Young Elite Talent Program. He currently serves as Vice President of the Chinese Society for Industrial and Applied Mathematics (CSIAM) and Chair of the CSIAM Youth Committee, and is on the editorial board of seven domestic and international journals, including TPAMI.

Faculty and students are welcome to attend!

Invited by: Kai Tu

School of Mathematical Sciences

April 22, 2026