Skip to content

Control Theories for Machine Learning (CT4ML)

Control Theories for Machine Learning

(CT4ML)

Why bother? It is time not only to ask what AI/ML can do for control but also ask what control (theories) can do for AI/ML. Control theory will enable us to rigorously analyze and synthesize AI/ML algorithms such that by bridging control theory and ML, AI/ML algorithms are both high-performing and provably stable, robust, and efficient.

Who cares? Researchers in general and Ph.D. students in particular who want to develop new ML algorithms based on rigorous analysis and design processes aiming “optimal yet robust machine learning”

Prerequisites? Control-1 (SISO); Control-2 (MIMO); Optimization (static and/or dynamic).

Textbook: n/a – should be developed

Instructor: Prof. YangQuan Chen, University California Merced, ychen53@ucmerced.edu

Control Theories for Machine Learning (CT4ML) Course Syllabus CT4ML .docx [16 hours, Jan. 2025 version]

News:

  • 7/5/2025. Dr. Chen's invited seminar. Control theory for machine learning (CT4ML): from integer order to fractional order ( 控制理论助力机器学习：从整数阶到分数阶) (Slides) at the 2025 BJTU AFC Workshop.
  • 6/15/2025. An old paper: Yuquan Chen, Yiheng Wei, YangQuan Chen. "Finite-time and Fixed-time Convergence in Continuous-time Optimization". [Submitted on 30 Sep 2021]  https://arxiv.org/abs/2109.15064
  • 3/27/2025. ASME/IEEE IDETC/CIE MESA FDTA2025 newly submitted paper. "A Survey on Finite-, Fixed-, and Prescribed-Time Convergent Optimization Algorithms: Control-Theoretic Perspectives" by Necdet Sinan Özbek, Osama F. Abdel Aal, YangQuan Chen
  • 3/27/2025. ASME/IEEE IDETC/CIE MESA FDTA2025 newly submitted paper.  "Tail-Index-Awareness in Fractional Order Stochastic Gradient Descent" by Mohammad Partohaghighi, Roummel Marcia, and YangQuan Chen
  • 3/17/2025. L-CSS+CDC25 submission. "Controlled Optimization with a Prescribed Finite-Time Convergence Using a Time Varying Feedback Gradient Flow" (6 pages) by Osama F. Abdel Aal, Necdet Sinan Özbek, Jairo Viola and YangQuan Chen [arXiv link: http://arxiv.org/abs/2503.13910] [IEEE L-CSS under review]
  • 3/17/2025. Prof. YangQuan Chen's talk on "CT4ML: Control Theory for Machine Learning Algorithm Analysis and Design - An Overview"  then with CDC24 plenary talk watching. 5-6pm. Youtube links. PPTs.
  • Jan. 2025. 16 hours short course on "Control Theory for Machine Learning (CT4ML)" developed and trial offered. Box folder.
     

People:

Our Papers:

Resources:

 

Created and last updated by Prof. YangQuan Chen 2/12/2025.