BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Computational Neuroscience Journal Club - Laurence Aitchison (CBL)
DTSTART:20190205T160000Z
DTEND:20190205T170000Z
UID:TALK119794@talks.cam.ac.uk
CONTACT:Rodrigo Echeveste
DESCRIPTION:Laurence Aitchison will be presenting:\n\n• Task representat
 ions in neural networks trained to perform many cognitive tasks\n\n• Gua
 ngyu Robert Yang\, Madhura R. Joglekar\, H. Francis Song\, William T. News
 ome & Xiao-Jing Wang\n\n• Nature Neuroscience 2019\n\n• https://www.na
 ture.com/articles/s41593-018-0310-2\n\nAbstract: The brain has the ability
  to flexibly perform many tasks\, but the underlying mechanism cannot be e
 lucidated in traditional experimental and modeling studies designed for on
 e task at a time. Here\, we trained single network models to perform 20 co
 gnitive tasks that depend on working memory\, decision making\, categoriza
 tion\, and inhibitory control. We found that after training\, recurrent un
 its can develop into clusters that are functionally specialized for differ
 ent cognitive processes\, and we introduce a simple yet effective measure 
 to quantify relationships between single-unit neural representations of ta
 sks. Learning often gives rise to compositionality of task representations
 \, a critical feature for cognitive flexibility\, whereby one task can be 
 performed by recombining instructions for other tasks. Finally\, networks 
 developed mixed task selectivity similar to recorded prefrontal neurons af
 ter learning multiple tasks sequentially with a continual-learning techniq
 ue. This work provides a computational platform to investigate neural repr
 esentations of many cognitive tasks.\n\n
LOCATION:Cambridge University Engineering Department\, CBL\, BE4-38 (http:
 //learning.eng.cam.ac.uk/Public/Directions)
END:VEVENT
END:VCALENDAR