BEGIN:VCALENDAR
VERSION:2.0
PRODID:IEEE vTools.Events//EN
CALSCALE:GREGORIAN
BEGIN:VTIMEZONE
TZID:America/New_York
BEGIN:DAYLIGHT
DTSTART:20260308T030000
TZOFFSETFROM:-0500
TZOFFSETTO:-0400
RRULE:FREQ=YEARLY;BYDAY=2SU;BYMONTH=3
TZNAME:EDT
END:DAYLIGHT
BEGIN:STANDARD
DTSTART:20261101T010000
TZOFFSETFROM:-0400
TZOFFSETTO:-0500
RRULE:FREQ=YEARLY;BYDAY=1SU;BYMONTH=11
TZNAME:EST
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20260515T170118Z
UID:0553F4AD-2923-492B-AD55-A1E9B7D3F95D
DTSTART;TZID=America/New_York:20260515T120000
DTEND;TZID=America/New_York:20260515T130000
DESCRIPTION:In this presentation\, we summarize our recent work on multi-ag
 ent reinforcement learning (MARL) under nonstationarity\, motivated by the
  fact that practical multi-agent systems are intrinsically nonstationary b
 ecause multiple agents learn and adapt concurrently. We frame three fundam
 ental challenges that hinder reliable deployment: (i) learning effective p
 olicies under continually changing peer behaviors\, (ii) reasoning about t
 he long-run evolution of other agents’ learning dynamics rather than onl
 y short-horizon updates\, and (iii) providing robustness guarantees under 
 perturbations arising from both the environment and peer learning agents. 
 We then present three technical thrusts to address the respective challeng
 es. First\, we introduce meta multi-agent reinforcement learning\, in whic
 h MARL is modeled as a Markov chain over joint policies\; this perspective
  enables a meta policy-gradient method that accounts not only for each age
 nt’s self-learning gradient but also an explicit peer-learning gradient\
 , thereby shaping how other agents adapt in the future. Second\, we addres
 s long-term influence in MARL by moving beyond few-step opponent modeling 
 and targeting limiting interaction outcomes as learning proceeds\, which i
 s essential when other agents continue to adapt indefinitely. Third\, we p
 resent ROMAX\, a framework for certifiably robust deep MARL via convex rel
 axation\, designed to improve resilience against nonstationary adversaries
  and cyber-physical perturbations by incorporating formal structure into r
 obustness-oriented objectives. Across representative matrix games and coup
 led multi-agent control benchmarks\, the presented methods demonstrate imp
 roved performance\, adaptation\, and robustness relative to standard MARL 
 baselines.\n\nSpeaker(s): Chuangchuang Sun\, \n\nVirtual: https://events.v
 tools.ieee.org/m/557218
LOCATION:Virtual: https://events.vtools.ieee.org/m/557218
ORGANIZER:guoxian2012@gmail.com
SEQUENCE:23
SUMMARY:IEEE PES Schenectady Chapter (Virtual) Lecture on Multiagent Reinfo
 rcement Learning under Nonstationarity
URL;VALUE=URI:https://events.vtools.ieee.org/m/557218
X-ALT-DESC:Description: <br /><p class="MsoNormal">In this presentation\, w
 e summarize our recent work on multi-agent reinforcement learning (MARL) u
 nder nonstationarity\, motivated by the fact that practical multi-agent sy
 stems are intrinsically nonstationary because multiple agents learn and ad
 apt concurrently. We frame three fundamental challenges that hinder reliab
 le deployment: (i) learning effective policies under continually changing 
 peer behaviors\, (ii) reasoning about the long-run evolution of other agen
 ts&rsquo\; learning dynamics rather than only short-horizon updates\, and 
 (iii) providing robustness guarantees under perturbations arising from bot
 h the environment and peer learning agents. We then present three technica
 l thrusts to address the respective challenges. First\, we introduce meta 
 multi-agent reinforcement learning\, in which MARL is modeled as a Markov 
 chain over joint policies\; this perspective enables a meta policy-gradien
 t method that accounts not only for each agent&rsquo\;s self-learning grad
 ient but also an explicit peer-learning gradient\, thereby shaping how oth
 er agents adapt in the future. Second\, we address long-term influence in 
 MARL by moving beyond few-step opponent modeling and targeting limiting in
 teraction outcomes as learning proceeds\, which is essential when other ag
 ents continue to adapt indefinitely. Third\, we present ROMAX\, a framewor
 k for certifiably robust deep MARL via convex relaxation\, designed to imp
 rove resilience against nonstationary adversaries and cyber-physical pertu
 rbations by incorporating formal structure into robustness-oriented object
 ives. Across representative matrix games and coupled multi-agent control b
 enchmarks\, the presented methods demonstrate improved performance\, adapt
 ation\, and robustness relative to standard MARL baselines.</p>
END:VEVENT
END:VCALENDAR