BEGIN:VCALENDAR
VERSION:2.0
PRODID:icalendar-ruby
CALSCALE:GREGORIAN
METHOD:PUBLISH
BEGIN:VTIMEZONE
TZID:Europe/Vienna
BEGIN:DAYLIGHT
DTSTART:20240331T030000
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
RRULE:FREQ=YEARLY;BYDAY=-1SU;BYMONTH=3
TZNAME:CEST
END:DAYLIGHT
BEGIN:STANDARD
DTSTART:20231029T020000
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
RRULE:FREQ=YEARLY;BYDAY=-1SU;BYMONTH=10
TZNAME:CET
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20260415T200228Z
UID:1710496800@ist.ac.at
DTSTART:20240315T110000
DTEND:20240315T120000
DESCRIPTION:Speaker: Marco Mondelli\nhosted by Dan Alistarh\nAbstract: Toda
 y\, we are at the center of a revolution in information technology\, with 
 data being the most valuable commodity. Exploiting this exploding number o
 f data sets requires to address complex inference problems\, spanning diff
 erent fields and arising in a variety of applications from engineering and
  the natural sciences. In particular\, in machine learning\, given a model
  for the observations\, the focus is on (i) how many samples convey suffic
 ient information to perform a certain task\, (ii) how to design low-comple
 xity algorithms that optimally utilize these samples\, and (iii) what are 
 the properties of the solution found by such algorithms. While the high di
 mensionality of the problem is often regarded as a ‘curse’ which makes
  it challenging to answer these questions\, in the talk I will discuss how
  the high dimensionality is in fact the key ingredient that allows analyti
 cal tractability. Specifically\, in the first part\, I will focus on gradi
 ent descent algorithms optimizing over-parameterized deep neural networks.
  In the second part\, I will move towards fundamental models\, such as gen
 eralized linear models and principal component analysis\, discussing how t
 o achieve the Bayes-optimal limits of inference. 
LOCATION:Raiffeisen Lecture Hall\, ISTA
ORGANIZER:maria.ariassutil@ist.ac.at
SUMMARY:Marco Mondelli: Inference in High Dimensions: Information Theoretic
  Limits and Efficient Algorithms
URL:https://talks-calendar.ista.ac.at/events/4630
END:VEVENT
END:VCALENDAR