BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Distributed distributional codes for learning successor features i
 n partially observable environments - Eszter Vértes\, Gatsby Unit\, Unive
 rsity College London
DTSTART:20200227T130000Z
DTEND:20200227T140000Z
UID:TALK140293@talks.cam.ac.uk
CONTACT:Microsoft Research Cambridge Talks Admins
DESCRIPTION:Animals need to devise strategies to maximise returns while in
 teracting with their environment based on incoming noisy sensory observati
 ons. Task-relevant states\, such as the agent's location within an environ
 ment or the presence of a predator\, are often not directly observable but
  must be inferred using available sensory information. Successor represent
 ations (SR) have been proposed as a middle-ground between model-based and 
 model-free reinforcement learning strategies\, allowing for fast value com
 putation and rapid adaptation to changes in the reward function or goal lo
 cations. Indeed\, recent studies suggest that features of neural responses
  are consistent with the SR framework. However\, it is not clear how such 
 representations might be learned and computed in partially observed\, nois
 y environments. Here\, we introduce a neurally plausible model using distr
 ibutional successor features\, which builds on the distributed distributio
 nal code for the representation and computation of uncertainty\, and which
  allows for efficient value function computation in partially observed env
 ironments via the successor representation. We show that distributional su
 ccessor features can support reinforcement learning in noisy environments 
 in which direct learning of successful policies is infeasible.
LOCATION:Auditorium\, Microsoft Research Ltd\, 21 Station Road\, Cambridge
 \, CB1 2FB
END:VEVENT
END:VCALENDAR