BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//wp-events-plugin.com//7.2.3.1//EN
TZID:Europe/Paris
X-WR-TIMEZONE:Europe/Paris
BEGIN:VEVENT
UID:6436@i2m.univ-amu.fr
DTSTART;TZID=Europe/Paris:20210423T143000
DTEND;TZID=Europe/Paris:20210423T153000
DTSTAMP:20241120T201432Z
URL:https://www.i2m.univ-amu.fr/evenements/keops-kernel-operations-on-the-
 gpu-with-autodiff-without-memory-overflows/
SUMMARY:Paul Escande (I2M\, CNRS\, Marseille): KeOps: Kernel operations on 
 the GPU\, with autodiff\, without memory overflows
DESCRIPTION:Paul Escande: The KeOps library is developed by Benjamin Charli
 er\, Jean Feydy and Joan Alexis Glaunès.\nIt lets you compute reductions 
 of large arrays whose entries are given by a mathematical formula or a neu
 ral network. It combines efficient C++ routines with an automatic differen
 tiation engine and can be used with Python (NumPy\, PyTorch)\, Matlab and 
 R.\nIt is perfectly suited to the computation of kernel matrix-vector prod
 ucts\, K-nearest neighbors queries\, N-body interactions\, point cloud con
 volutions and the associated gradients. Crucially\, it performs well even 
 when the corresponding kernel or distance matrices do not fit into the RAM
  or GPU memory. Compared with a PyTorch GPU baseline\, KeOps provides a x1
 0-x100 speed-up on a wide range of geometric applications\, from kernel me
 thods to geometric deep learning.\nDISCLAIMER: I am NOT one of the authors
  of the library\, I happen to use it and thought it could be of interest f
 or many researchers. My knowledge of the library is thus not complete but 
 I believe this seminar would provide a glimpse of its operation.\nhttps://
 www.kernel-operations.io/keops/index.html\n
CATEGORIES:Séminaire,Signal et Apprentissage
END:VEVENT
BEGIN:VTIMEZONE
TZID:Europe/Paris
X-LIC-LOCATION:Europe/Paris
BEGIN:DAYLIGHT
DTSTART:20210328T030000
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
TZNAME:CEST
END:DAYLIGHT
END:VTIMEZONE
END:VCALENDAR