Speaker: YangQuan Chen, Dept. of Mechanical and Aerospace Engineering, University of California, Merced

Time : 15:00 - 16.00 CET (Rome/Paris)

Hosted at: SISSA, International School of Advanced Studies, Trieste, Italy

Zoom :  A zoom meeitng link will appear here, one hour before the talk

Organizers : Pavan Pranjivan Mehta* (pavan.mehta@sissa.it) and Arran Fernandez** (arran.fernandez@emu.edu.tr)

* SISSA, International School of Advanced Studies, Italy

** Eastern Mediterranean University, Northern Cyprus

Keywords: roughness, roughness informed, machine learning, federated learning, fractional calculus, fractal calculus

Abstract: This talk advocates that machine learning ought to be not only "physics-informed" but also "complexity-informed" so that smarter machine learning becomes possible. After introducing the triangle of "inverse power law - complexity - fractional calculus'' we show that both fractal calculus and fractional calculus are mathematical vehicle for tail behavior characterization therefore the exponential law (EL, integer order calculus), stretched exponential law (SEL, fractal calculus) and inverse power law (IPL, fractional calculus) are in a unified view. We then show that roughness concept is important in machine learning when loss landscape roughness is considered. Roughness in the sense of statistics, manifold, geometrical etc. can be quantified by using fractal and fractional calculi. Machine learning algorithms that are aware of roughness and are informed by roughness can perform much better than those conventional machine learning algorithms that do not respect the complexity or roughness information. The take home message is simple: AI/machine learning and fractional calculus should marry.

Biography: Prof. YangQuan Chen is with the Dept. of Mechanical and Aerospace Engineering at the University of California Merced. He received his B.S. from the University of Science and Technology of Beijing, M.S. from Beijing Institute of Technology, and Ph.D. from Nanyang Technological University Singapore. His research interests include smart mechatronics for sustainability, smart control engineering via digital twins, small multi-UAV based cooperative multi-spectral ``remote sensing", applied fractional calculus in complex system controls, modeling, signal processing, and machine learning; distributed measurement and control of distributed parameter systems with mobile actuator and sensor networks. He authored many papers, editorials, patents, research monographs and textbooks. His Google Scholar H index = 107 total citations = 61256. His latest books with CRC Press are Fractional Calculus for Skeptics (I), (II).

Bibliography

[1] Mohammad Partohaghighi, Roummel Marcia, YangQuan Chen. “Roughness-Informed Machine Learning – A Call for Fractal and Fractional Calculi.” Journal of Information and Intelligence. (Sept. 2025) [DOI: https://doi.org/10.1016/j.jiixd.2025.09.001
[2] Mohammad Partohaghighi, Roummel Marcia, YangQuan Chen. “Roughness-Informed Machine Unlearning.” In Proc. of the 2025 ICFDA, Algiers, Algeria (to appear Dec. 2025)
[3] Fractional Calculus for Federated Learning (FC4FL) webpage: https://mechatronics.ucmerced.edu/fc4fl
[4] Mohammad Partohaghighi. “Smart Federated Learning and Unlearning: Theoretical Framework, Benchmark Studies, and Applications,” Ph.D. dissertation, EECS, UC Merced, May 2026.
[5] Optimization, Big Data, and AI/ML: A section of Fractal and Fractional (ISSN 2504-3110). https://www.mdpi.com/journal/fractalfract/sections/Optimization_Big_Data...