BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Data-driven models of neural and behavioural learning - N Alex Cay
 co Gajic\, École Normale Supérieure
DTSTART:20250121T150000Z
DTEND:20250121T163000Z
UID:TALK226852@talks.cam.ac.uk
CONTACT:Daniel Kornai
DESCRIPTION:Learning and adaptation play a central role in interacting wit
 h a dynamic environment. Neuroscience experiments have classically focused
  on how individual brain regions perform simplified tasks. However\, recen
 t technological advances have rapidly enabled the monitoring of large popu
 lations of neurons over many days\, across multiple brain regions\, and du
 ring increasingly complex behaviors. Yet even with such data within our re
 ach\, we still lack the theoretical and quantitative tools necessary to in
 fer the fundamental principles guiding learning in the brain.\n\nIn this t
 alk I will present several of our latest efforts to bridge this gap. First
 \, by building state-dependent statistical models\, we demonstrate that co
 mplex locomotor behaviours can be disentangled into a twofold learning pro
 cess combining discrete and continuous aspects that are both refined over 
 learning. Second\, we propose a theoretical framework for how different mo
 tor regions (the motor cortex and cerebellum) could coordinate as a distri
 buted learning system. Finally\, I will present our recent development of 
 tensor-based dimensionality reduction methods to track how neural dynamics
  change as learning unfolds. Together\, our work aims to develop interpret
 able data-driven models to understand and link learning dynamics across ne
 ural and behavioural scales.
LOCATION:CBL Seminar Room\, Engineering Department\, 4th floor Baker build
 ing
END:VEVENT
END:VCALENDAR