BEGIN:VCALENDAR
PRODID:-//eluceo/ical//2.0/EN
VERSION:2.0
CALSCALE:GREGORIAN
BEGIN:VEVENT
UID:www.tcs.tifr.res.in/event/1662
DTSTAMP:20251230T044235Z
SUMMARY:Algorithmically Testing Sub-Gaussianity of High Dimensional Distrib
 utions
DESCRIPTION:Speaker: Ankit Pensia (CARNEGIE MELLON UNIVERSITY\, USA)\n\nAbs
 tract: \nSub-Gaussian distributions play a central role in statistics\, pr
 obability\, and computer science. A distribution is termed sub-Gaussian i
 f all of its univariate projections have tails that decay faster than a Ga
 ussian. In algorithmic statistics\, a central task is the following: give
 n samples from a distribution\, decide whether the underlying distribution
  is sub-Gaussian or heavy-tailed. In high dimensions\, this is a challeng
 ing problem because sub-Gaussianity requires all univariate projections to
  be light-tailed\, which seemingly necessitates a brute-force search over 
 directions to verify. In this talk\, I will describe a structural propert
 y of sub-Gaussian distributions that leads to computationally efficient al
 gorithms for a wide range of statistical problems\, e.g.\, clustering\, ro
 bust estimation\, and more. Based on joint work with Ilias Diakonikolas\,
  Sam Hopkins\, and Stefan Tiegel.\n \nShort Bio: Ankit Pensia is an assis
 tant professor in the Department of Statistics and Data Science at Carnegi
 e Mellon University. Previously\, he was a research fellow at the Simons I
 nstitute for the Theory of Computing and a Herman Goldstine Postdoctoral F
 ellow at IBM Research. He obtained his PhD in Computer Science from the Un
 iversity of Wisconsin-Madison. His current research interests include algo
 rithmic robust statistics\, high-dimensional probability\, distribution te
 sting\, and algorithmic stability.\n \n
URL:https://www.tcs.tifr.res.in/web/events/1662
DTSTART;TZID=Asia/Kolkata:20260107T110000
DTEND;TZID=Asia/Kolkata:20260107T120000
LOCATION:A-201 (STCS Seminar Room)
END:VEVENT
END:VCALENDAR