BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:Quantum maximum likelihood prediction via Hilbert space embeddings
DTSTART:20261005T160000
DTEND:20261005T170000
DTSTAMP:20261002T231646Z
UID:2969bafc1072b3a9ed3bb3feff924355035a1930bedbbd46319356c5
CATEGORIES:Conferences - Seminars
DESCRIPTION:Prof. Nir Weinberger\, Technion\nMaximum likelihood prediction
  plays a major (hidden) role in modern large language models via the next 
 token prediction task. In this talk\, we propose and study a simplified ve
 rsion\, based on an i.i.d. data model. The model is based on a quantum ma
 ximum likelihood predictor\, which is obtained by embedding empirical dist
 ributions into quantum states and minimizing quantum relative entropy over
  a prescribed model class.\nWe interpret this predictor via quantum revers
 e information projection and a quantum Pythagorean theorem\, under structu
 ral assumptions such as unitary invariance and closure under pinching. We 
 extend this theorem to non-self-adjoint mixture families in finite dimensi
 ons and establish a related infinite-dimensional inequality under addition
 al regularity conditions. Finally\, we present non-asymptotic guarantees\
 , including convergence rates and concentration inequalities in trace norm
  and quantum relative entropy. \nJoint work with Sreejith Sreekumar. \n
 \nBio : Nir Weinberger is an Associate Professor at the The Viterbi Facul
 ty of Electrical and Computer Engineering\, Technion - Israel Institute of
 Technology. Previously\, from 2017 to 2018 he was a post-doctoral fellow a
 t Tel-Aviv University\, and from 2018-2020 he was a Technion-MIT post-doct
 oral fellow at the Massachusetts Institute of Technology\, Cambridge\, MA\
 , USA. He has received the B.Sc. and M.Sc. degrees from Tel-Aviv Universit
 y\, Tel-Aviv\, Israel\, in 2006 and 2009\, respectively\, and his Ph.D.deg
 ree in 2017\, from the Technion\, Israel Institute of Technology.
LOCATION:BC 129 https://plan.epfl.ch/?room==BC%20129
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR