BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//pretalx//pretalx.com//juliacon2021//speaker//UHXALN
BEGIN:VEVENT
UID:pretalx-juliacon2021-EVR3HZ@pretalx.com
DTSTART:20210729T163000Z
DTEND:20210729T170000Z
DESCRIPTION:We present InvertibleNetworks.jl\, an open-source package for i
 nvertible neural networks and normalizing flows using memory-efficient bac
 kpropagation. InvertibleNetworks.jl uses manually implement gradients to t
 ake advantage of the invertibility of building blocks\, which allows for s
 caling to large-scale problem sizes. We present the architecture and featu
 res of the library and demonstrate its application to a variety of problem
 s ranging from loop unrolling to uncertainty quantification.
DTSTAMP:20260511T115351Z
LOCATION:Green
SUMMARY:InvertibleNetworks.jl - Memory efficient deep learning in Julia - P
 hilipp A. Witte\, Mathias Louboutin\, Ali Siahkoohi\, Felix J. Herrmann\, 
 Gabrio Rizzuti\, Bas Peters
URL:https://pretalx.com/juliacon2021/talk/EVR3HZ/
END:VEVENT
END:VCALENDAR