Towards Safe AI for Cyber-Physical Systems
Title: | Towards Safe AI for Cyber-Physical Systems |
Speaker: | Professor Partha Roop (Department of Electrical, Computer, and Software Engineering, University of Auckland) |
Time: | 2024年11月29日 (周五) 15:00 |
Venue: | 线下:中国科学院软件园区5号楼四层 第三会议室 线上:腾讯会议 693-138-482 |
Abstract: | In this talk I will discuss challenges for AI-based control of CPS using the case study of a cardiac pacemaker. Then I will present the following approaches to designing AI-enabled controllers, which are safety certifiable. Specifically, I will elaborate on the following recent attempts from our research group: — A formal approach for ensuring the safety of reinforcement learning enabled controllers (RV 2024, Istanbul) — A formal approach for compositional neural networks inspired by synchronous programming (MEMOCODE 2018 – Beijing and DATE 2020 –Germany) — An approach for the verification of spiking neural networks (IEEE Transactions on Computers 2023) |
Bio: | Partha Roop is a Professor and Associate Dean International in the Faculty of Engineering, The University of Auckland, New Zealand. His research interests are mainly in formal methods, especially in CPS applications in biomedical engineering, digital health, and intelligent transportation systems. Partha has a PhD from UNSW Sydney and Master of Technology from Indian Institute of Technology Kharagpur, both in Computer Science and Engineering. |