BEGIN:VCALENDAR
PRODID:-//eluceo/ical//2.0/EN
VERSION:2.0
CALSCALE:GREGORIAN
BEGIN:VEVENT
UID:www.tcs.tifr.res.in/event/1773
DTSTAMP:20260910T044744Z
SUMMARY:Trustworthy Reinforcement Learning: From Formal Specifications to L
 earning Algorithms
DESCRIPTION:Speaker: Ashutosh Trivedi (University of Colorado Boulder)\n\nA
 bstract: \nReinforcement learning provides a powerful framework for learni
 ng decision-making strategies from interaction\, but specifying what an ag
 ent should learn is often as challenging as designing the learning algorit
 hm itself. A scalar reward can be an inadequate description of objectives 
 that involve sequences of events\, long-term requirements\, safety constra
 ints\, or other temporal properties.\nIn this talk\, I will discuss how id
 eas from formal methods can help address this problem. Temporal logic\, au
 tomata\, and related specification formalisms provide precise ways to desc
 ribe desired behavior\, while reinforcement-learning algorithms provide me
 chanisms for learning how to achieve such objectives from experience. I wi
 ll describe ways in which these viewpoints can be connected so that richer
  specifications can be incorporated into learning algorithms without losin
 g the ability to reason about their behavior.\nI will also discuss some of
  our recent work on reinforcement learning with formally specified objecti
 ves and on obtaining statistical and formal guarantees for learned policie
 s. More broadly\, the talk will illustrate how classical ideas from algori
 thms\, automata\, and verification can contribute to the design of learnin
 g systems that are easier to specify\, analyze\, and trust.\nShort bio: As
 hutosh Trivedi is an Associate Professor of Computer Science at the Univer
 sity of Colorado Boulder. His research lies at the intersection of formal 
 methods\, reinforcement learning\, and trustworthy artificial intelligence
 \, with applications to cyber-physical and safety-critical systems. His wo
 rk develops mathematical foundations and algorithms for specifying\, learn
 ing\, and verifying the behavior of learning-enabled systems.\n
URL:https://www.tcs.tifr.res.in/web/events/1773
DTSTART;TZID=Asia/Kolkata:20261210T160000
DTEND;TZID=Asia/Kolkata:20261210T170000
LOCATION:Homi Bhabha Auditorium
END:VEVENT
END:VCALENDAR