BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Deep Neural Networks and Multigrid Methods - Jinchao Xu (Pennsylva
 nia State University)
DTSTART:20191030T140500Z
DTEND:20191030T150500Z
UID:TALK134185@talks.cam.ac.uk
CONTACT:INI IT
DESCRIPTION:In this talk\, I will first give an introduction to several mo
 dels and algorithms from two different fields: (1) machine learning\, incl
 uding logistic regression\, support vector machine and deep neural network
 s\, and (2) numerical PDEs\, including finite element and multigrid method
 s.&nbsp\; I will then explore mathematical relationships between these mod
 els and algorithms and demonstrate how such relationships can be used to u
 nderstand\, study and improve the model structures\, mathematical properti
 es and relevant training algorithms for deep neural networks. In particula
 r\, I will demonstrate how a new convolutional neural network known as MgN
 et\, can be derived by making very minor modifications of a classic geomet
 ric multigrid method for the Poisson equation and then explore the theoret
 ical and practical potentials of MgNet.   <br><br><br><br><br>
LOCATION:Seminar Room 2\, Newton Institute
END:VEVENT
END:VCALENDAR