BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Controlling Behavioral Diversity in Multi-Agent Reinforcement Lear
 ning - Matteo Bettini (University of Cambridge)
DTSTART:20241126T130000Z
DTEND:20241126T140000Z
UID:TALK223090@talks.cam.ac.uk
CONTACT:Mateja Jamnik
DESCRIPTION:Diversity has been shown to be key to collective intelligence 
 in natural systems. Despite this\, current Multi-Agent Reinforcement Learn
 ing (MARL) approaches enforce behavioral homogeneity (to boost efficiency)
  or blindly promote behavioral diversity via intrinsic rewards or addition
 al loss functions\, effectively changing the learning objective and lackin
 g a principled measure for it. In this context\, the present work deals wi
 th the question of how to control the diversity of a multi-agent system.  
 We introduce Diversity Control (DiCo)\, a method able to control diversity
  to an exact value of a given metric by representing policies as the sum o
 f a parameter-shared component and dynamically scaled per-agent components
 . By applying constraints directly to the policy architecture\, DiCo leave
 s the learning objective unchanged\, enabling its applicability to any act
 or-critic MARL algorithm. We theoretically prove that DiCo achieves the de
 sired diversity\, and we provide several experiments\, both in cooperative
  and competitive tasks\, that show how DiCo can be employed as a novel par
 adigm to increase performance and sample efficiency in MARL\, as well as l
 ead to the emergence of novel diverse policies. Multimedia results are ava
 ilable on the "project’s website":https://sites.google.com/view/dico-mar
 l.\n\n\n"You can also join us on Zoom":https://cam-ac-uk.zoom.us/j/8340033
 5522?pwd=LkjYvMOvVpMbabOV1MVTm8QU6DrGN7.1
LOCATION:Lecture Theatre 2\, Computer Laboratory\, William Gates Building
END:VEVENT
END:VCALENDAR