BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Robust nonnegative matrix factorization with the beta-divergence a
 nd applications in imaging - Cédric Févotte\, Institut de Recherche en I
 nformatique de Toulouse (IRIT)
DTSTART:20250422T100000Z
DTEND:20250422T110000Z
UID:TALK229834@talks.cam.ac.uk
CONTACT:Kimberly Cole
DESCRIPTION:Data is often available in matrix form\, in which columns are 
 samples\, and processing of such data often entails finding an approximate
  factorization of the matrix into two factors. The first factor (the “di
 ctionary”) yields recurring patterns characteristic of the data. The sec
 ond factor (“the activation matrix”) describes in which proportions ea
 ch data sample is made of these patterns. Nonnegative matrix factorization
  (NMF) is a popular unsupervised learning technique for analysing data wit
 h nonnegative values\, with applications in many areas such as in text in
 formation retrieval\, recommender systems\, audio signal processing\, and 
 hyperspectral imaging. In a first part\, I will give a short tutorial pres
 entation about NMF for data processing\, with a focus on majorization-mini
 mization algorithms for NMF with the beta-divergence\, a continuous family
  of loss functions that takes the quadratic loss\, KL divergence and Itaku
 ra-Saito divergence as special cases. Then\, I will present applications f
 or hyperspectral unmixing in remote sensing and factor analysis in dynamic
  positron emission tomography\, introducing robust variants of NMF that ac
 count for outliers\, nonlinear phenomena or specific binding.\n\nReference
 s\nC. Févotte\, J. Idier. Algorithms for nonnegative matrix factorization
  with the beta-divergence. Neural computation\, 2011.\nC. Févotte\, N. Do
 bigeon. Nonlinear hyperspectral unmixing with robust nonnegative matrix fa
 ctorization. IEEE Transactions on Image Processing\, 2015.\nY. C. Cavalcan
 ti\, T. Oberlin\, N. Dobigeon\, C. Févotte\, S. Stute\, M. J. Ribeiro\, C
 . Tauber. Factor analysis of dynamic PET images: beyond Gaussian noise. IE
 EE Transactions on Medical Imaging\, 2019.\nA. Marmin\, H. Goulart\, C. F
 évotte. Joint majorization-minimization for nonnegative matrix factorizat
 ion with the beta-divergence. IEEE Transactions on Signal Processing\, 202
 3.\n\n\nBio:\n\nCédric Févotte is a CNRS research director with Institut
  de Recherche en Informatique de Toulouse (IRIT). Previously\, he was a CN
 RS researcher at Laboratoire Lagrange (Nice\, 2013-2016) & Télécom Paris
 Tech (2007-2013)\, a research engineer at Mist-Technologies (the startup t
 hat became Audionamix\, 2006-2007) and a postdoc at University of Cambridg
 e (2003-2006). He holds MEng and PhD degrees in EECS from École Centrale 
 de Nantes. His research interests concern statistical signal processing an
 d machine learning\, with particular interests in matrix factorization\, i
 nverse problems\, source separation and recommender systems. Selected dist
 inctions: IEEE Fellow (2022)\, ERC Consolidator Grant (2016-2022)\, IEEE S
 ignal Processing Society Sustained Impact Paper Award (2023)\, IEEE Signal
  Processing Society Best Paper Award (2014).\n\n
LOCATION:LT6\, Baker Building\, CUED
END:VEVENT
END:VCALENDAR