BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:IC Talk: How hard is it to learn from data generated by a probabil
 istic model?
DTSTART:20191105T111500
DTEND:20191105T123000
DTSTAMP:20260408T012958Z
UID:2a561f172177a925a739c5c44b15ee3af71b4bdf68528bacba6c8755
CATEGORIES:Conferences - Seminars
DESCRIPTION:By: Florent Krzakala - Sorbonne University\nVideo of his talk\
 n\nAbstract:\nIn many modern applications\, large instances of NP-hard pro
 blems are routinely solved. While this should be impossible from the worst
 -case perspective\, the behaviour of algorithms on relevant “typical” 
 data is often not captured by the worst case. This motivates the following
  question: How hard is it to learn from data generated by a probabilistic 
 model?\n\nI will consider several prototypical generative models and learn
 ing tasks\, e.g. detecting communities in graphs\, performing compressed s
 ensing and phase retrieval\, or learning a rule with multi-layer neural ne
 tworks\, for which it is possible to obtain closed-form answers to this qu
 estion. Interestingly\, in a wide variety of such problems we find sharp c
 hanges --phase transitions-- in statistical and computational performance 
 as some parameters are changed. In particular\, there exists a region of p
 arameters where the underlying statistical problem is information-theoreti
 cally possible to solve yet no efficient algorithm is known\, rendering th
 e problem essentially unsolvable in practice.\n\nIn the final part of the 
 talk\, I will finally build of the fact that realistic structured data can
  often be captured via generative adversarial networks or variational auto
 -encoders\, and using them can lead to drastic improvements. I will presen
 t a methodology that enables us to extend the analysis of the statistical 
 and computational performance to such data generated by random multi-layer
  neural networks.\n\nBio:\nFlorent Krzakala is a professor at Sorbonne Uni
 versité and a Researcher at Ecole Normale Superieure in Paris. His resear
 ch interests include Statistical Physics\, Machine Learning\, Statistics\,
  Computer Science and Computational Optics. He leads the SPHINX “Statist
 ical PHysics of INformation eXtraction” team in Ecole Normale in Paris\,
  and is the holder of the CFM-ENS Datascience chair and of a PRAIRIE Insti
 tute chair. He is also the funder and scientific advisor of the startup Li
 ghton.\n\nMore information
LOCATION:BC 420 https://plan.epfl.ch/?room==BC%20420
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR