BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//132.216.98.100//NONSGML kigkonsult.se iCalcreator 2.20.4//
BEGIN:VEVENT
UID:20260908T222601EDT-7856vW1upz@132.216.98.100
DTSTAMP:20260909T022601Z
DESCRIPTION:Title: Depth Degeneracy and Vanishing. Angles for Random Deep N
 eural Networks\n\nAbstract: Stacking many layers to create truly *deep* ne
 ural networks is arguably what has led to the recent explosion in AI. Howe
 ver\, many properties of deep neural networks are not yet understood. One 
 such mystery is the depth degeneracy phenomenon: the deeper you make your 
 network\, the closer your network is to a constant function on initializat
 ion. In this talk\, we examine the evolution of the angle between two inpu
 ts to a ReLU neural network as a function of the number of layers. By usin
 g combinatorial expansions\, we find precise formulas for how fast this an
 gle goes to zero as depth increases. The formulas are given in terms of th
 e mixed moments of correlated Gaussians passed through the ReLU function. 
 We also find a surprising combinatorial connection between these mixed mom
 ents and the Bessel numbers.\n\nIn person or by Zoom link: https://mcgill.
 zoom.us/j/89737173009?pwd=UzlwZkVPK0RnYXk4VGM2aXo4V3Q2QT09\n\n \n\n \n
DTSTART:20230223T163000Z
DTEND:20230223T173000Z
LOCATION:Room 1214\, Burnside Hall\, CA\, QC\, Montreal\, H3A 0B9\, 805 rue
  Sherbrooke Ouest
SUMMARY:Mihai Nica (Guelph)
URL:https://www.mcgill.ca/mathstat/channels/event/mihai-nica-guelph-346257
END:VEVENT
END:VCALENDAR