Date: Tuesday, August 20, 2024

Time: 11:00 am – 12:30pm EST

Zoom Registration link:

https://us06web.zoom.us/meeting/register/tZcudumgqzIrH9EmD_nN3lSFfGtcCscvsiMF#/

(After you register, a confirmation email with calendar information will be sent out)

Recording: TBA

Talk Title: Learning with Limited Resources: Optimizing Neural Networks for Training, Adaptation, and Deployment

Abstract:

In this talk, we explore a few cutting-edge methodologies for optimizing neural networks in resource-constrained environments. To this end, we address key challenges in neural architecture search, dynamic adaptation, on-device training, and vision transformer adaptation, while focusing on significantly reducing the computational and memory requirements and maintaining high performance for various IoT applications. By optimizing for efficiency, these methods pave the way towards more scalable, adaptable, and sustainable AI solutions, thus facilitating broader deployment across diverse, resource-limited settings. Finally, we also contextualize these advancements within the broader landscape of AI/ML research, highlighting their implications for federated learning, sustainable AI, and real-time applications.

Bio:

Radu Marculescu is a Professor and the Laura Jennings Turner Chair in Engineering in the Department of Electrical and Computer Engineering at The University of Texas at Austin. Between 2000-2019, he was a Professor in the Electrical and Computer Engineering department at Carnegie Mellon University. His current research focuses on developing AI/ML algorithms and tools for system design and optimization for computer vision, bioimaging, and IoT applications. He has received the 2019 IEEE Computer Society Edward J. McCluskey Technical Achievement Award, for seminal contributions to the science of network on chip design, analysis, and optimization. He has also received the 2020 ESWEEK Test-of-Time Award from The International Conference on Hardware/Software Co-Design and System Synthesis (CODES). He is an IEEE Fellow and an ACM Fellow.

Date: Wednesday, April 24, 2024

Time: 11:00 am – 12:30pm EST

Zoom Registration link:

https://us06web.zoom.us/meeting/register/tZIsdu6qpjwrH9R0hTu0VK6jZkRggQg4NofS

Recording: TBA

Talk Title: Verifiably Safe Deep Reinforcement Learning-Enabled Control

Abstract:

My talk consists of two parts. Part one is on lessons on engineering higher-impact research. I created a 3-credit graduate class, Improving Your Research Skills, for graduate students. This class consists of invited talks from outstanding colleagues, e.g., Xue Liu and Jiawei Han, on their insights and students’ presentations on how they apply what they learned to their own research.  I will share a few lessons.  Part two is on my current research on verifiably safe Deep Reinforcement Learning enabled control.  The certification of a safety-critical cyber-physical system (CPS), e.g., flight control, requires (i) an analytical model of system behaviors with proof of safety properties and (ii) verification and validation of the software implementation.  Current Artificial Intelligence (AI) architectures cannot satisfy these two requirements. The root cause is using a non-causal statistical method to approximate causal deterministic relations from Newtonian physics. For example, the Markovian decision process is the foundation of the widely used Deep Reinforcement Learning (DRL) in control applications. Some of the RL control commands will inevitably be inconsistent with the law of physics. Hallucination is in the “DNA” of DRL.  Fortunately, the laws of physics enable us to achieve 1) system-level verifiable safety, 2) reducing AI hallucination probability, and 3) more explainable AI controller behavior. I will give an overview of our approach towards this goal.

Bio:

Lui Sha graduated with a Ph.D. from CMU in 1985. He worked at the Software Engineering Institute from 1986 to 1998. He joined UIUC in 1998 as a full professor. He is Donald B. Gillies Chair Professor of the Computer Science Department, and Daniel C. Drucker, Eminent Faculty at UIUC’s College of Engineering. He is a fellow of IEEE and ACM and a recipient of the IEEE Simon Ramo Medal for exceptional achievement in systems engineering and systems science. He was a member of the National Academic of Science’s Committee on Certifiably Dependable Software Systems and a member of the NASA Advisory Council. He also received the Test of Time and Influential Paper awards.

He led the research, development, and transition to practice on real-time and embedded computing technologies, cited as a significant accomplishment in the selected accomplishment section of the 1992 National Academy of Science’s report, “A Broader Agenda for Computer Science and Engineering” (P.193). He led a comprehensive revision of IEEE standards on real-time computing, which have since become the best practice in real-time computing systems. It has been widely used in real-time systems such as airplanes, robots, cars, ships, trains, and medical devices.

His recent research includes 1) Physics Model Regulated DNN and 2) Medical Guidance Systems for Acute diseases, such as Sepsis, which claims ~250,000 lives per year and is the third leading cause of death in the USA alone.

Research Impact Examples

• “Transformation of real-time computing practice from an ad hoc process to an engineering process based on analytic methods.” IEEE Fellow citation.
• “Technical leadership and contributions to fundamental theory, practice and standardization for engineering real-time systems.”  IEEE Simon Ramo Medal citation.
• Rescuing Mars Pathfinder: “The Mars Pathfinder mission was widely proclaimed as “flawless” in the early days after its July 4th, 1997 landing on the Martian surface.  …  But a few days into the mission, not long after Pathfinder started gathering meteorological data, the spacecraft began experiencing total system resets… Once diagnosed, it was clear to the JPL engineers that using priority inheritance would prevent the resets they were seeing. …No more system resets occurred. … When was the last time you saw a room of people cheer a group of computer science theorists for their significant practical contribution to advancing human knowledge? 🙂 It was quite a moment.” http://catless.ncl.  ac.uk/Risks/ 19.49.html
• Global Positioning Satellite in orbit software upgrade: “The navigation payload software for the next block of Global Positioning System upgrades recently completed testing. …  This design would have been difficult or impossible before the development of rate monotonic theory“, L.  Doyle, and J. Elzey “Successful Use of Rate Monotonic Theory on A Formidable Real-Time System, technical report, p.1, ITT, Aerospace Communication Division, 1993.
• International Space Station: “Through the development of Rate Monotonic Scheduling, we now have a system that will allow [Space Station] Freedom’s computers to budget their time, to choose between a variety of tasks, and decide not only which one to do first but how much time to spend in the process,” Aaron Cohen, Deputy Administrator of NASA, October 1992 (p. 3), Charting The Future: Challenges and Promises Ahead of Space Exploration.

Date: Wednesday, Feb 21st, 2024

Time: 11:00 am – 12:30pm EST

Zoom Registration link: https://us06web.zoom.us/meeting/register/tZUvcu6gqDIsGNwGvze0l–PcnPp9yq7YBMH

Recording: link

Talk Title: Toward High-Assurance Learning-Enabled Cyber-Physical Systems

Abstract:

Learning-Enabled Cyber-Physical Systems are becoming increasingly essential to society. Many advances have been made in the last decade in constructing autonomous Cyber-Physical Systems (CPS) as evidenced by the proliferation of unmanned systems in air, ground, and sea. These advances have been driven by innovations in several areas, such as computing platform technologies, control theory, design methods and tools, machine learning, modeling, and simulation technologies, among others. In particular, machine learning provides a potentially revolutionary way for extracting functionalities needed for higher-level autonomy. Unfortunately, our current lack of understanding of when and how machine learning works makes it challenging to provide guarantees for learning-enable components in safety critical systems. Despite this limitation, given the impressive experimental results of machine learning, researchers have quickly incorporated learning in perception-action loops even in driverless cars and aerial vehicles, where the safety requirements are very high. This has resulted in unreliable behavior and public failures (e.g., Tesla and drone crashes, Uber running a red light) that may lead to loss of trust in autonomy.

There are many challenges in developing autonomous Learning-Enabled CPS that are safe and secure. This talk will present assurance problems, challenges, and techniques.  They include safety verification of closed-loop systems with neural network components, confidence estimation and composition at runtime, assumption monitoring and checking of CPS, Out-of-Distribution (OOD) detection and adversarial digital and physical attack detection, and CPS checkpointing and recovery.

Bio:

Insup Lee is Cecilia Fitler Moore Professor of Computer and Information Science and Director of PRECISE Center since 2008 at the University of Pennsylvania.  He also holds a secondary appointment in the Department of Electrical and Systems Engineering. His research interests include cyber-physical systems (CPS), real-time systems, embedded systems, high-confidence medical device systems, formal methods and tools, run-time verification, and adversarial learning.  The theme of his research activities has been to assure and improve the correctness, safety, and timeliness of life-critical embedded systems. His papers received the nine best conference paper awards. Recently, he has been working in Internet of Medical Things, security of cyber physical systems, and safe autonomy.

He has served on many program committees, chaired many international conferences and workshops, and served on various steering and advisory committees of technical societies. He is founding co-Editor-in-Chief of ACM Transactions on Computing for Healthcare (HEALTH, 2018) and was founding co-Editor-in-Chief of KIISE Journal of Computing Science and Engineering (JCSE, 2007). He has also served on the editorial boards on the several scientific journals, including Journal of ACM, ACM Transactions on Cyber-Physical Systems, IEEE Transactions on Computers, Formal Methods in System Design, and Real-Time Systems Journal. He was founding Co-Director of Penn Health Tech (2017-2020). He was Chair of ACM Special Interest Group on Embedded Systems (SIGBED, 2015-2019) and Chair of IEEE TC on Real-Time Systems (TCRTS, 2003-2004). He was a member of Technical Advisory Group (TAG) of President’s Council of Advisors on Science and Technology (PCAST) Networking and Information Technology (2006-2007).  He was a member of the National Research Council’s committee on 21st Century Cyber-Physical Systems Education (2014-2015). He received IEEE TC-RTS Outstanding Technical Achievement and Leadership Award in 2008. He received an appreciation award from Ministry of Science, IT and Future Planning, South Korea in 2013. His work received the Runtime Verification (RV) Test-of-Time award in 2019. He received ACM SIGBED Distinguished Leadership Award in 2022 and IEEE Technical Committee on Cyber-Physical Systems (TCCPS) Distinguished Leadership Award in 2023. He is ACM fellow, IEEE fellow and AAAS fellow.

Date: Thursday, Sep 21st, 2023

Time: 11:00 am – 12:30pm EST

Zoom Registration link: https://us06web.zoom.us/meeting/register/tZUvcu6gqDIsGNwGvze0l–PcnPp9yq7YBMH

Recording: link

Talk Title: Towards Ambient Intelligence for Healthcare: A CPS Perspective

Abstract: Ambient Intelligence has been a goal for more than 20 years. Are we getting close? What if we focus ambient intelligence on smart healthcare, are we getting close? What role does CPS play in ambient intelligence? This talk is motivated by these questions. Various challenges, research directions, and research results from my group’s work will be used to (partially) address these themes for smart healthcare. The talk includes discussions on the ambient intelligence vision, the role of CPS, cognitive assistants on wearables, solutions supporting mental health, and lessons learned from real deployments. There is also a brief discussion on two key ML/CPS challenges: the need for robust models and dealing with uncertainties due to the environment and human behaviors.

Bio: Professor John A. Stankovic is the BP America Professor in the Computer Science Department at the University of Virginia and Ex-Director of the (CPS) Link Lab. He is a Fellow of both the IEEE and the ACM. He has been awarded an Honorary Doctorate from the University of York for his work on real-time systems. In 2022, he was elected to the Virginia Academy of Science, Engineering, and Medicine. He won the IEEE Real-Time Systems Technical Committee’s Award for Outstanding Technical Contributions and Leadership.  He also received the IEEE Technical Committee on Distributed Processing’s Distinguished Achievement Award (inaugural winner), and the IEEE TC on CPS’s Technical Achievement Award. He has two test-of-time paper awards. Stankovic has an h-index of 124 and over 66,300 citations. Prof. Stankovic received his PhD from Brown University.

Date: Thursday, June 15th, 2023

Time: 11:00 am – 12:30pm EST

Zoom Registration link: https://us06web.zoom.us/meeting/register/tZcpfu6rqTMvGdfcZdzy5lclTZXkM0eElT-u#/registration

Recording: https://youtu.be/OA_GknXKe4g

Talk Title: Deterministic Concurrency in Cyber-Physical Systems

Abstract: POPULAR frameworks based on publish-and-subscribe, service-oriented architectures, and actor networks have simplified the development of parallel and distributed software. Most of these, however, are intrinsically non-deterministic. This talk will examine some of the risks that this introduces to applications, particularly cyber-physical applications. I will then introduce an international collaboration that aims to fix the problem through a polyglot coordination language called Lingua Franca (LF). I will show that deterministic concurrency does not automatically imply a cost in performance. Finally, I will focus on how LF enables navigating the fundamental and unavoidable tradeoff between consistency and availability in distributed systems.

Bio: Edward A. Lee has been working on embedded software systems for more than 40 years. He is Professor of the Graduate School in EECS. His research is focused on cyber-physical systems. He leads the open-source software project Lingua Franca and previously Ptolemy II, is a coauthor of textbooks on embedded systems, signals and systems, digital communications, and philosophical and social implications of technology. His current research is focused on a polyglot coordination language for distributed real-time systems called Lingua Franca that combines features of discrete-event modeling, synchronous languages, and actors.

He is a Fellow of the IEEE, was an NSF Presidential Young Investigator, won the 1997 Frederick Emmons Terman Award for Engineering Education, received the 2016 Outstanding Technical Achievement and Leadership Award from the IEEE Technical Committee on Real-Time Systems (TCRTS), the 2018 Berkeley Citation, the 2019 IEEE Technical Committee on Cyber-Physical Systems (TCCPS) Technical Achievement Award, and the 2022 European Design and Automation Association (EDAA) Achievement Award.