BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:Dualities in Neural Networks
DTSTART:20210412T140000
DTEND:20210412T150000
DTSTAMP:20260429T231814Z
UID:ab38657c166ae150620e983c76a70c9d249d44e86d9dc3ba911b3472
CATEGORIES:Conferences - Seminars
DESCRIPTION:James Halverson (Northeastern)\nDualities give different descr
 iptions of the same physical system that can allow for progress in one dua
 lity frame that is not possible in another. In this talk I'll introduce du
 alities in neural networks. In the first case in appropriate large-N limit
 s\, neural networks are functions drawn from Gaussian processes\, allowing
  a perturbative non-Gaussian approach at finite N\, akin to quantum field 
 theory. In this limit the parameter space complexity formally goes to infi
 nity\, while the function space perspective analogous to effective field t
 heory becomes more tractable. Yet these are dual descriptions of the same 
 system\, and I will exploit duality to deduce symmetries of the function-s
 pace action even when it is not explicitly known. Time permitting\, I will
  also demonstrate analogs of so-called IR dualities in neural networks\, l
 eading to notions of universality via the study of correlation functions.\
 n 
LOCATION:zoom https://epfl.zoom.us/j/83233601454?pwd=emhTYjBjR3E4YzFrMVVpN
 nhxQ05KZz09
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR