BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:Multiple and Hierarchical Universality
DTSTART:20230208T110000
DTEND:20230208T120000
DTSTAMP:20260407T091331Z
UID:93d5a9a0275f9d049a95fe8158878c40a1bf6e31e117b0683462c69c
CATEGORIES:Conferences - Seminars
DESCRIPTION:Prof. Meir Feder    University of Tel Aviv\nUniversal codin
 g\, prediction and learning usually consider the case where the data gener
 ating mechanism is unknown or non-existent\, and the goal of the universal
  scheme is to compete with the best hypothesis from a given hypothesis cla
 ss\, either on the average or in a worst-case scenario. Multiple universal
 ity considers the case where the hypothesis class is also unknown: there a
 re several hypothesis classes with possibly different complexities. In hie
 rarchical universality\, the simpler classes are nested within more comple
 x classes. The main challenge is to correctly define the universality crit
 erion so that the extra "regret" for not knowing the class is monitored. W
 e propose possible definitions and derive their min-max optimal solutions.
  Interestingly\, the proposed solutions can be used to obtain Elias codes 
 for universal representation of the integers. We also utilize this approac
 h for variable-memory Markov models (unifilar models)\, presenting a new i
 nterpretation for the bound over the regret of the celebrated context-tree
  weighting algorithm and propose a 3-part code that (slightly) out-perform
 s it. Finally\, we conjecture that multiple universality with its non-unif
 orm regret can be used in other "overparameterized" model classes includin
 g deep neural networks.\nJoint work with Yaniv Fogel\n--------------------
 --------------------------------------------------------------------------
 -----\nMeir Feder received the Sc.D degree in Electrical Engineering and O
 cean Engineering in 1987 from the Massachusetts Institute of Technology (M
 IT) and the Woods Hole Oceanographic Institution (WHOI). After being a Res
 earch Associate and a Lecturer in MIT\, he joined in 1990 the School of El
 ectrical Engineering\, Tel-Aviv University\, where he is now the Jokel Cha
 ir Professor and the head of the newly established Tel-Aviv university cen
 ter for Artificial intelligence and Data science (TAD). He is also a Visit
 ing Professor with the Department of EECS\, MIT. Parallel to his academic 
 career\, he is closely involved with the high-tech industry. He founded 5 
 companies\, among them are Peach Networks that developed an interactive TV
  solution (Acq: MSFT) and Amimon that provided the highest quality\, robus
 t and no delay wireless high-definition A/V connectivity (Acq:LON.VTC). Re
 cently\, with his renewed interest in machine learning and AI\, he cofound
 ed Run:ai\, a virtualization\, orchestration\, and acceleration platform f
 or AI infrastructure. He is also an active angel investor and serves on th
 e board/advisory board of several US and Israeli companies. Prof. Feder re
 ceived several academic and professional awards including the IEEE Informa
 tion Theory Society best paper award for his work on universal prediction\
 , the "creative thinking" award of the Israeli Defense Forces and the Rese
 arch Prize of the Israeli Electronic Industry\, awarded by the President o
 f Israel. For the development of Amimon's chip-set\, that uses a unique MI
 MO implementation of joint source-channel coding for wireless video transm
 ission he received the 2020 Scientific and Engineering Award of the Academ
 y of Motion Picture Arts and Sciences (Oscar) and was announced as the pri
 ncipal inventor of the technology that attained the 73rd Engineering Emmy 
 Award of the Television Academy.
LOCATION:INR 113 https://plan.epfl.ch/?room==INR%20113
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR