BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//CERN//INDICO//EN
BEGIN:VEVENT
SUMMARY:Huddle on Learning and Inference from Structured Data: Universalit
y\, Correlations and Beyond | (smr 3855)
DTSTART;VALUE=DATE-TIME:20230710T060000Z
DTEND;VALUE=DATE-TIME:20230721T200000Z
DTSTAMP;VALUE=DATE-TIME:20240613T094143Z
UID:indico-event-10189@ictp.it
DESCRIPTION:An ICTP meeting in person\n In the continuation of the worksho
p\, the huddle will focus on the modelisation and statistical analysis of
large structured data sets as appearing in modern signal processing and ma
chine learning. Recurrent questions will be: what are ``good’’ models
of high-dimensional data which are realistic enough while remaining analyt
ically tractable\, and what are their universality properties?\n \n Many
modern problems (e.g. compressed sensing\, community detection\, PCA and
tensor decomposition) seek to infer some latent signal from high-dimension
al noisy data. A theoretical analysis is often challenging due to the subt
le correlations and structured dependencies among the observed features. R
emarkably\, many of these systems exhibit universal statistics ie.\, simil
ar properties as a surrogate random system. There has been substantial rec
ent progress at the intersection of statistical physics\, statistics\, pro
bability and machine learning in rigorously establishing these empirical o
bservations\, and these properties have been critically exploited for stat
istical learning.\n This huddle will focus on these recent interdisciplina
ry investigations\, with a view towards discovering new connections among
the diverse approaches to these problems of common interest.\nTopics:\n H
igh-dimensional statistics and inference\n \n Statistical learning\n \n
Models of structured data\n \n Universality\n \n Statistical mechanics\n
\n\n//indico.ictp.it/event/10189/
LOCATION:ICTP Giambiagi Lecture Hall (AGH)
URL://indico.ictp.it/event/10189/
END:VEVENT
END:VCALENDAR