BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//132.216.98.100//NONSGML kigkonsult.se iCalcreator 2.20.4//
BEGIN:VEVENT
UID:20250707T045117EDT-3064VjsUND@132.216.98.100
DTSTAMP:20250707T085117Z
DESCRIPTION: \n\nTITLE: Modern ``Non''-Problems in Optimization: Applicatio
 ns to Statistics and Machine Learning\n\nAbstract:  We have witnessed a lo
 t of exciting development of data science in recent years. From the perspe
 ctive of optimization\, many modern data-science problems involve some bas
 ic ``non’’-properties that lack systematic treatment by the current approa
 ches for the sake of the computation convenience. These non-properties inc
 lude the coupling of the non-convexity\, non-differentiability and non-det
 erminism. In this talk\, we present rigorous computational methods for sol
 ving two typical non-problems: the piecewise linear regression and the fee
 d-forward deep neural network. The algorithmic framework is an integration
  of the first order non-convex majorization-minimization method and the se
 cond order non-smooth Newton methods. Numerical experiments demonstrate th
 e effectiveness of our proposed approach. Contrary to existing methods for
  solving non-problems which provide at best very weak guarantees on the co
 mputed solutions obtained in practical implementation\, our rigorous mathe
 matical treatment aims to understand properties of these computed solution
 s with reference to both the empirical and the population risk minimizatio
 ns. This is based on joint work with Jong-Shi Pang\, Bodhisattva Sen and Z
 iyu He.\n
DTSTART:20190125T210000Z
DTEND:20190125T220000Z
LOCATION:Room 1104\, Burnside Hall\, CA\, QC\, Montreal\, H3A 0B9\, 805 rue
  Sherbrooke Ouest
SUMMARY:Ying Cui- University of South California
URL:/mathstat/channels/event/ying-cui-university-south
 -california-293620
END:VEVENT
END:VCALENDAR