BEGIN:VCALENDAR
PRODID:-//eluceo/ical//2.0/EN
VERSION:2.0
CALSCALE:GREGORIAN
BEGIN:VEVENT
UID:www.tcs.tifr.res.in/event/1616
DTSTAMP:20250911T063920Z
SUMMARY:Learning to Control Unknown Multi-Agent Systems
DESCRIPTION:Speaker: Siddharth Chandak (Stanford University)\n\nAbstract: \
 nLarge-scale multi-agent systems are often modeled as games\, where each p
 layer's reward depends on the joint actions of all agents. In strongly mon
 otone games\, players converge to a Nash equilibrium (NE) by optimizing th
 eir local objectives\, but such equilibria may not align with the global o
 bjective. We study two scenarios where a game manager\, with access only t
 o the global objective and limited control over utility parameters\, seeks
  to steer the system toward better equilibria.\nIn the first scenario\, th
 e controller adjusts linear coefficients in the players' utilities to impo
 se linear constraints on the equilibrium. We design a simple two-time-scal
 e stochastic approximation algorithm and show almost sure convergence and 
 a mean square error rate of near-$O(t^{-1/4})$ for the algorithm.In the se
 cond scenario\, the game manager has to choose among K discrete parameters
 . We propose a novel optimism-based bandit algorithm with additional terms
  to account for the distance from equilibrium\, and prove that this algori
 thm achieves a regret of O(log(T)).\nShort Bio:Siddharth Chandak is curren
 tly a Ph.D. candidate in Electrical Engineering at Stanford University\, U
 SA. He received his B.Tech. from IIT Bombay\, India\, in 2021\, where he w
 as awarded the President of India Gold Medal\, and his M.S. from Stanford 
 University in 2023. His research interests include game theory\, multi-age
 nt learning\, stochastic approximation\, and its applications in reinforce
 ment learning.\n
URL:https://www.tcs.tifr.res.in/web/events/1616
DTSTART;TZID=Asia/Kolkata:20250917T160000
DTEND;TZID=Asia/Kolkata:20250917T170000
LOCATION:A-201 (STCS Seminar Room)
END:VEVENT
END:VCALENDAR