BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//talks.cam.ac.uk//v3//EN
BEGIN:VTIMEZONE
TZID:Europe/London
BEGIN:DAYLIGHT
TZOFFSETFROM:+0000
TZOFFSETTO:+0100
TZNAME:BST
DTSTART:19700329T010000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=-1SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:+0100
TZOFFSETTO:+0000
TZNAME:GMT
DTSTART:19701025T020000
RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=-1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
CATEGORIES:Isaac Newton Institute Seminar Series
SUMMARY:Opening up the black box of score estimation - Sit
 an Chen (Harvard University)
DTSTART;TZID=Europe/London:20240719T110000
DTEND;TZID=Europe/London:20240719T120000
UID:TALK219070AThttp://talks.cam.ac.uk
URL:http://talks.cam.ac.uk/talk/index/219070
DESCRIPTION:In recent years there has been significant interes
 t in the theoretical foundations of diffusion gene
 rative modeling. One representative result in this
  line of work is that with an accurate estimate of
  the score function for the data distribution\, on
 e can approximately sample from virtually any boun
 ded distribution in polynomial time. In this talk 
 I will describe recent work on the missing piece l
 eft open by these works: when can we actually lear
 n an accurate estimate of the score from data? I w
 ill focus on two vignettes: (1) learning Gaussian 
 mixture models (GMMs)\, and (2) learning optimal e
 stimators for compressed sensing.For (1)\, I will 
 present an algorithm for score estimation based on
  piecewise polynomial regression\, yielding the fi
 rst quasipolynomial-time algorithm for learning ge
 neral mixtures of Gaussians with polylogarithmical
 ly many components. For (2)\, I will give the firs
 t rigorous learning guarantee for algorithm unroll
 ing\, proving that a certain unrolled network\, wh
 en trained on compressed sensing examples\, learns
  to compete with Bayes approximate message passing
 .Based on joint works with Aayush Karan\, Vasilis 
 Kontonis\, and Kulin Shah.\n&nbsp\;
LOCATION:External
CONTACT:
END:VEVENT
END:VCALENDAR