BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:Faster Kernel Matrix Algebra via Density Estimation
DTSTART:20230525T130000
DTEND:20230525T140000
DTSTAMP:20260507T173857Z
UID:b705dc52f958e9a59fe84af91dd6425e2232f117728a07c14b97245c
CATEGORIES:Conferences - Seminars
DESCRIPTION:Professor Piotr Indyk - MIT\nComputational Mathematics Seminar
  \n \nAbstract :\nKernel matrices\, as well as weighted graphs represente
 d by them\, are ubiquitous objects in machine learning\, statistics and ot
 her related fields. The main drawback of using kernel methods (learning an
 d inference using kernel matrices) is efficiency – given n input points\
 , most kernel-based algorithms need to materialize the full n × n kernel 
 matrix before performing any subsequent computation\, thus incurring\nΩ(n
 ^2) runtime. Breaking this quadratic barrier for various problems has ther
 efore\, been a subject of extensive research efforts.\nIn this talk I will
  present fast algorithms for computing basic properties of an nxn positive
  semidefinite kernel matrices K\, induced by n points x_1\,...\, x_n in R^
 d. This includes estimating the sum of kernel matrix entries\, computing i
 ts top eigenvalue and eigenvector\, spectral sparsification\, low-rank app
 roximation\, etc. The algorithms leverage the positive definiteness of the
  kernel matrix\, along with a recent line of work on efficient kernel dens
 ity estimation.\n 
LOCATION:GA 3 21 https://plan.epfl.ch/?room==GA%203%2021
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR