Intelligent Decision-Making Under Uncertainty: Beyond Probabilities
- RC Trust
Prof. Dr. Nils Jansen, Institute for Computing and Information Science (ICIS) at Radboud University, Nijmegen, The Netherlands
Abstract: This talk highlights our vision of foundational and application-driven research toward safety in artificial intelligence (AI). We take abroad stance on AI that combines machine learning, control theory, andformal methods, particularly formal verification. As part of thisresearch line, we study problems inspired by autonomous systems,planning in robotics, and direct industrial applications.We consider reinforcement learning as a specific machine learningtechnique, which generally learns to behave optimally via trial anderror.Consequently, and despite its massive success in the past years,reinforcement learning lacks mechanisms to ensure safe and correctbehavior. Formal verification is a research area that provides formalguarantees of a system’s correctness and safety based on rigorousmethods and precise specifications.Yet, fundamental challengesobstruct the effective application of verification to reinforcementlearning so far.Our main objective is to devise novel, data-driven verificationmethods that tightly integrate with reinforcement learning. Inparticular, we develop techniques that address real-world challengesto the safety of AI systems in general: Scalability, expressiveness,and robustness against the uncertainty that occurs when operating inthe real world. The overall goal is to advance the real-worlddeployment of reinforcement learning.
Prof. Dr. Nils Jansen
Bio: Nils Jansen is an associate professor with the Institutefor Computing and Information Science (ICIS) at Radboud University,Nijmegen, The Netherlands. He received his Ph.D. with distinction fromRWTH Aachen University, Germany, in 2015. Prior to Radboud University,he was a research associate at the University of Texas at Austin. Hisresearch is on intelligent decision-making under uncertainty, with afocus on formal reasoning about safety and dependability aspects inartificial intelligence (AI). He holds several grants in academic andindustrial settings, among them an ERC starting grant with the title:Data-Driven Verification and Learning Under Uncertainty (DEUCE).