BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//132.216.98.100//NONSGML kigkonsult.se iCalcreator 2.20.4//
BEGIN:VEVENT
UID:20260303T175834EST-72026lvsEb@132.216.98.100
DTSTAMP:20260303T225834Z
DESCRIPTION:Computing Under Extreme Connectivity Constraints in Artificial 
 and Biological Neural Networks\n\nWilten Nicola\, University of Calgary\n	T
 uesday March 10\, 12-1pm\n	Zoom Link: https://mcgill.zoom.us/j/87078928687
 \n	In Person: 550 Sherbrooke\, Room 189\n	\n	Abstract: Many existing models o
 f computation utilizing recurrent neural networks assume dense\, unconstra
 ined initial connectivity\, where any pair of neurons may be coupled to ge
 nerate the rich dynamics needed for learning complex temporal patterns. In
 spired by invertebrate circuits that often exhibit ring-like connectivity\
 , we show that computation can occur in ultra-sparse spiking and rate rese
 rvoirs that are initially coupled as simple unidirectional rings or other 
 ring structures. In contrast to standard recurrent networks\, the total nu
 mber of network parameters in these ring networks scales only linearly wit
 h network size\, while still producing rich feature sets. We demonstrate t
 hat such networks can successfully reproduce a range of dynamical systems 
 tasks\, including oscillations\, multi-stable switches\, and low-dimension
 al chaotic attractors. Our findings show that structured spatio-temporal d
 ynamics naturally arising from large ring topologies\, often observed in i
 nvertebrate circuits\, are a sufficient mechanism for learning different t
 ypes of attractors. This work was conducted with Dr. Afroditi Talidou at t
 he Hotchkiss Brain Institute.\n
DTSTART:20260310T160000Z
DTEND:20260310T170000Z
SUMMARY:QLS Seminar Series - Wilten Nicola
URL:/miam/channels/event/qls-seminar-series-wilten-nic
 ola-371466
END:VEVENT
END:VCALENDAR