BEGIN:VCALENDAR
VERSION:2.0
METHOD:PUBLISH
CALSCALE:GREGORIAN
PRODID:-//WordPress - MECv6.5.6//EN
X-ORIGINAL-URL:https://cmc.deusto.eus/
X-WR-CALNAME:cmc.deusto.eus
X-WR-CALDESC:DeustoCCM - Chair of Computational Mathematics at University of Deusto
REFRESH-INTERVAL;VALUE=DURATION:PT1H
X-PUBLISHED-TTL:PT1H
X-MS-OLK-FORCEINSPECTOROPEN:TRUE
BEGIN:VEVENT
CLASS:PUBLIC
UID:MEC-939ce5a14db795adecaaa63048a18d50@cmc.deusto.eus
DTSTART:20211022T083000Z
DTEND:20211022T093000Z
DTSTAMP:20251031T214200Z
CREATED:20251031
LAST-MODIFIED:20251031
PRIORITY:5
TRANSP:OPAQUE
SUMMARY:Analysis of gradient descent on wide two-layer ReLU neural networks
DESCRIPTION:Speaker: Dr. Lénaïc Chizat\nAffiliation: EPFL, École Polytechnique Fédérale de Lausanne (Switzerland)\nOrganized by: FAU DCN-AvH, Chair for Dynamics, Control and Numerics – Alexander von Humboldt Professorship at FAU Erlangen-Nürnberg (Germany)\nZoom meeting link\nMeeting ID:  615 4539 3381 | PIN: 304949\nAbstract. In this talk, we propose an analysis of gradient descent on wide two-layer ReLU neural networks that leads to sharp characterizations of the learned predictor. The main idea is to study the training dynamics when the width of the hidden layer goes to infinity, which is a Wasserstein gradient flow. While this dynamics evolves on a non-convex landscape, we show that for appropriate initializations, its limit, when it exists, is a global minimizer. We also study the implicit regularization of this algorithm when the objective is the unregularized logistic loss, which leads to a max-margin classifier in a certain functional space. We finally discuss what these results tell us about the generalization performance, and in particular how these models compare to kernel methods.\nThis event on LinkedIn\n
URL:https://cmc.deusto.eus/events-calendar/analysis-of-gradient-descent-on-wide-two-layer-relu-neural-networks/
CATEGORIES:FAU DCN-AvH Seminar,Seminar/Talk
ATTACH;FMTTYPE=image/png:https://cmc.deusto.eus/wp-content/uploads/FAUDCNAvH-seminar-22oct2021-lChizat.png
END:VEVENT
END:VCALENDAR