Chuan He

Chuan He is an Assistant Professor in the Department of Mathematics at Linköping University, Sweden, affiliated with the Division of Applied Mathematics (TIMA) and the Wallenberg AI, Autonomous Systems and Software Program (WASP). His research bridges continuous optimization and machine learning, focusing on algorithmic efficiency and theoretical foundations.

Education:

• Ph.D. in Industrial and Systems Engineering, University of Minnesota, USA (2019–2023)
• B.S. in School of Mathematical Sciences, Xiamen University, China (2015–2019)

His research interests include deep learning, decentralized and large-scale optimization, high-order methods, and applications in healthcare, scientific computing, and engineering. He develops algorithms with strong theoretical guarantees, particularly in nonconvex optimization settings. His work emphasizes improving the speed, reliability, and scalability of machine learning training processes.

His recent publications focus on Newton-CG based methods, augmented Lagrangian techniques, and federated learning under constraints. These contributions span top journals in operations research, optimization, and machine learning, reflecting a strong trend toward integrating second-order optimization with practical machine learning challenges.

Scientific Awards: No awards explicitly mentioned in the text.

Chuan He advises and collaborates within the WASP Mathematics research environment, particularly in the 'Optimisation for machine learning' group. He previously held a postdoctoral position at the University of Minnesota under Professor Ju Sun. His research is supported through institutional affiliations with WASP and Linköping University. He actively contributes to the academic community through conference presentations at INFORMS, SIAM, and NeurIPS workshops.

He is involved in the 'Optimisation for machine learning' research group at MAI, which aims to develop more efficient and theoretically sound algorithms for machine learning. The group focuses on replacing heuristic methods with principled approaches, reducing computational costs in training models.