BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:Zoltan Szabo - Consistency of Orlicz Random Fourier Features
DTSTART:20190923T111500
DTEND:20190923T120000
DTSTAMP:20260916T194622Z
UID:c6bdee427f6601a9b2cbe0ff4658922fb87e569b6103f6a241f2c141
CATEGORIES:Conferences - Seminars
DESCRIPTION:Zoltán Szabó (http://www.cmap.polytechnique.fr/~zoltan.szabo
 /) is a Research Associate Professor at the Center of Applied Mathematics 
 (CMAP)\, École Polytechnique\, France. His main research interests are ke
 rnel methods\, information theory\, randomized algorithms and their applic
 ations. He serves/served as an Area Chair at ICML (2019\, 2018\, 2017)\, A
 ISTATS (2020\, 2019\, 2018\, 2017)\, NeurIPS (2018)\, UAI (2020\, 2017\, 2
 016) and IJCAI (2019)\, a Senior Area Chair at NeurIPS (2019)\, the Progra
 m Chair of the Data Science Summer School (DS^3-2019\, 2018\, 2017)\, and 
 he is the moderator of statistical machine learning (stat.ML) on arXiv.\nK
 ernel techniques provide highly flexible tools with successful application
 s at virtually all sub-fields of machine learning and statistics. The rand
 om Fourier feature approach (RFF) is probably the most widely-applied and 
 popular idea to combine this representational power of kernels with comput
 ational efficiency\; it won the 10-year test-of-time award at NIPS-2017. W
 hile the RFF technique is typically used in case of tasks expressed via fu
 nction values (such as kernel ridge regression)\, in numerous applications
  taking into account high-order derivatives turns out to be beneficial\; e
 xamples include nonlinear feature selection or fitting infinite-dimensiona
 l exponential family distributions. Despite its practical success\, the th
 eoretical understanding of RFFs in case of derivatives is rather limited. 
 In this talk\, I will show how a finite alpha-exponential Orlicz norm assu
 mption allows one to get consistent RFF approximations in case of high-ord
 er derivatives\, covering for example the popular inverse multiquadric (al
 pha = 1) or the Gaussian kernel (alpha = 2). This is a joint work with Lin
 da Chamakh and Emmanuel Gobet.
LOCATION:BC 329 https://plan.epfl.ch/?room==BC%20329
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR