BEGIN:VCALENDAR
PRODID:-//eluceo/ical//2.0/EN
VERSION:2.0
CALSCALE:GREGORIAN
BEGIN:VEVENT
UID:www.tcs.tifr.res.in/event/1627
DTSTAMP:20251008T094225Z
SUMMARY:Convex Gaussian Min-Max Theorem for Precise Error Analysis of Regul
 arized Regression Problems in High Dimensions
DESCRIPTION:Speaker: Tirthankar Adhikari (TIFR)\n\nAbstract: \nConvex regul
 arized optimization methods such as the LASSO and Group-LASSO are central 
 to modern signal processing and high-dimensional regression problems. In t
 heir influential work\, Thrampoulidis\, Oymak\, and Hassibi introduced the
  Convex Gaussian Min-Max Theorem (CGMT)\, an elegant framework for preci
 se asymptotic error analysis of a wide class of convex regression problem
 s under Gaussian measurement models. Building on Gordon's Gaussian min-max
  theorem and ideas of Stojnic\, they showed that a complicated Primary Op
 timization (PO) problem can be replaced by a simpler\, analytically tract
 able Auxiliary Optimization (AO) problem whose solution precisely predict
 s key quantities such as the normalized squared error and optimal regulari
 zation parameters. Beyond its specific applications to linear and structur
 ed regression\, the CGMT provides a powerful conceptual bridge between hig
 h-dimensional probability and convex optimization\, enabling sharp asympto
 tic performance characterizations through the remarkably elegant PO to AO
  reduction.\n
URL:https://www.tcs.tifr.res.in/web/events/1627
DTSTART;TZID=Asia/Kolkata:20251010T160000
DTEND;TZID=Asia/Kolkata:20251010T170000
LOCATION:A-201 (STCS Seminar Room)
END:VEVENT
END:VCALENDAR