BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//132.216.98.100//NONSGML kigkonsult.se iCalcreator 2.20.4//
BEGIN:VEVENT
UID:20260831T103605EDT-2477O1i9cu@132.216.98.100
DTSTAMP:20260831T143605Z
DESCRIPTION:Title: Universal Approximation Theorems\n\nAbstract: The univer
 sal approximation theorem established the density of specific families of 
 neural networks in the space of continuous functions and in certain Bochne
 r-Lebesgue spaces\, defined between any two Euclidean spaces. We extend an
 d refine this result by proving that there exist dense neural network arch
 itectures on a larger class of function spaces and that these architecture
 s may be written down using only a small number of functions. Refinements 
 of the classical results of Hornik 1989 are also obtained. We prove that u
 pon appropriately randomly selecting the neural networks architecture's ac
 tivation function we may still obtain a dense set of neural networks\, wit
 h positive probability. This last result is used to overcome the difficult
 y of appropriately selecting an activation function in more exotic archite
 ctures.\n
DTSTART:20191016T193000Z
DTEND:20191016T203000Z
LOCATION:Room LB 921-4\, CA\, Seminar Statistique Concordia
SUMMARY:Anastasis Kratsios\, ETH Zurich
URL:https://www.mcgill.ca/mathstat/channels/event/anastasis-kratsios-eth-zu
 rich-301625
END:VEVENT
END:VCALENDAR