Title: Formal Methods Enhanced Deep Learning for Cyber-Physical Systems
Date: February 25th, 2021 at 10:00am
* The above link works for those affiliated with the University of Pittsburgh. If you are not affiliated with the University of Pittsburgh, please email Heidi Davis at firstname.lastname@example.org to obtain the meeting information.
Presenter: Meiyi Ma
AI-powered Cyber-Physical Systems (AI-CPS) form the basis of emerging and future smart services, improve quality of life, and bring advances in many critical areas, including smart cities and health care. A central problem is how to build reliable AI-CPS facing significant new challenges due to the increasing complexity, integration, and environmental uncertainties. In this talk, I will present my approaches to developing rigorous and robust models for reliable AI-CPS by integrating formal methods and deep learning. First, I will give a brief overview of an AI-CPS decision-support system incorporating integrative solutions to improve its capability and reliability. This system is to be deployed in Newark City, connecting to the Mayor’s dashboard for conflict detection and resolution among smart services. Second, I will describe two of the key formal methods enhanced deep learning techniques from the system. The first technique increases the robustness of sequential prediction by incorporating formal specification and verification to the learning process, which is a powerful method to guide the network to follow critical CPS model properties. The second technique creates a novel specification logic to measure the uncertainty in deep learning with a confidence guarantee. It then applies the logic-based criteria to automatically calibrate the uncertainty estimation model for deep learning. I will also show the generalizability of these techniques for various neural networks and applications. I will conclude my talk with a brief discussion of my future research directions on further integrating formal methods and deep learning towards explainable, verifiable, and reliable AI-CPS addressing a broad spectrum of social and technical challenges.
Meiyi Ma is a Ph.D. candidate in the Department of Computer Science at the University of Virginia, working with Prof. John A. Stankovic and Prof. Lu Feng. Her research interest lies at the intersection of Machine learning, Formal Methods, and Cyber-Physical Systems. Specifically, her work develops rigorous and robust AI by integrating formal methods and machine learning, and applies new integrative solutions to build safe and reliable Cyber-Physical Systems, with a focus on smart city and healthcare applications. Meiyi’s research has been published in top-tier machine learning and cyber-physical systems conferences and journals. She has received multiple research awards, including the EECS Rising Star at UC Berkeley, the CDAC Rising Star in Data Science, the UVa Link Lab Outstanding Graduate Research Award and the Best Master Thesis Award. On professional services, she is serving as the information director for ACM Transactions on Computing for Healthcare. She also has served as an organizing committee member for several international workshops and a reviewer for multiple conferences and journals. For more information: http://www.cs.virginia.edu/~mm5tk
Host: Stephen Lee