Mahyar Fazlyab

Mahyar 

Postdoctoral Researcher

Electrical and Systems Engineering

University of Pennsylvania

Philadelphia, PA

Email: mahyarfa at seas dot upenn dot edu

Location: 407B, 3401 Walnut St. Philadelphia, PA 19104

About

I am a postdoctoral researcher in the Electrical and Systems Engineering (ESE) department at the University of Pennsylvania, working with Prof. Manfred Morari and Prof. George J. Pappas. Prior to that, I obtained my Ph.D. in ESE from UPenn in 2018 under the supervision of Prof. Victor M. Preciado. During my Ph.D. I also worked closely with Prof. Alejandro Ribeiro.

Research Summary

Artificial Intelligence (AI), and Deep Neural Networks at its forefront, have made groundbreaking advances in various industries, from self-driving cars to automated healthcare. However, there are still many fundamental questions and issues to address before we can deploy AI-driven technologies in the real world. In particular:

  • How can we provide certificates of stability, safety, and robustness for learning-enabled closed-loop systems in a scalable and modular fashion?

  • How can we make neural networks more robust, explainable, and data-efficient?

  • How can we design scalable optimization algorithms with multiple performance criteria such as robustness and adaptation in uncertain and dynamic environments?

My research aims at developing a rigorous understanding of the interplay between machine learning, control, and optimization and how they can complement each other in addressing the challenges described above. I am interested in bringing the strong theoretical guarantees of control theory to machine learning, the intelligence of machine learning to optimization, and scalable modular techniques of optimization to both, focusing on machine learning for optimization (e.g., speeding up combinatorial optimization problems), data-driven robust optimization in the context of learning and system identification, optimization over machine learning models (e.g., neural network verification, robust deep learning), and automated synthesis of novel (distributed) optimization algorithms for non-convex problems arising in machine learning applications.

Recent News

  • March 23rd 2020: Invited Seminar, Department of Electrical and Computer Engineering, University of Southern California (USC).

  • March 9th 2020: Invited Seminar, Department of Mechanical Engineering, Massachusetts Institute of Technology (MIT).

  • February 20th 2020: Invited Seminar, Mathematical Institute for Data Science (MINDS), Johns Hopkins University.

  • February 17th 2020: Invited Seminar, School of Computing, Informatics, and Decision Systems Engineering, Arizona State University (ASU).

  • For a complete list of news, click here.