BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//pretalx//pretalx.com//juliacon2024//speaker//XN93M7
BEGIN:VTIMEZONE
TZID:Europe/Amsterdam
BEGIN:DAYLIGHT
DTSTART:20230711T000000
TZNAME:CEST
TZOFFSETFROM:+0200
TZOFFSETTO:+0200
END:DAYLIGHT
BEGIN:STANDARD
DTSTART:20231029T030000
RDATE:20241027T030000
TZNAME:CET
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
END:STANDARD
BEGIN:DAYLIGHT
DTSTART:20240331T030000
RDATE:20250330T030000
TZNAME:CEST
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
END:DAYLIGHT
END:VTIMEZONE
BEGIN:VEVENT
SUMMARY:My Experience With Deep Learning Research in Julia. - Romeo Valent
 in
DTSTART;TZID=Europe/Amsterdam:20240710T105600
DTEND;TZID=Europe/Amsterdam:20240710T105900
DTSTAMP:20260809T104232Z
UID:pretalx-juliacon2024-NVFAVL@pretalx.com
DESCRIPTION:I spent about one year leveraging Julia for DL research in Com
 puter Vision\, including using `Flux.jl`\, `FastAI.jl`\, `Metalhead.jl`\, 
 and loading python models through `PyChainCall.jl`. Happy to discuss my ex
 perience.
LOCATION:Method (1.5)
URL:https://pretalx.com/juliacon2024/talk/NVFAVL/
END:VEVENT
BEGIN:VEVENT
SUMMARY:KSVD.jl: A case study in performance optimization. - Romeo Valenti
 n
DTSTART;TZID=Europe/Amsterdam:20240710T163000
DTEND;TZID=Europe/Amsterdam:20240710T164000
DTSTAMP:20260809T104232Z
UID:pretalx-juliacon2024-V8VGK9@pretalx.com
DESCRIPTION:We present `KSVD.jl`\, an extremely fast implementation of the
  K-SVD algorithm\, including extensive single-core optimizations\, shared-
 state multithreading\, pipelined GPU offloading\, and an optional distribu
 ted executor. With this implementation\, we are able to outperform existin
 g numpy-based implementations by ~100x and scale to datasets with millions
  of samples.
LOCATION:Function (4.1)
URL:https://pretalx.com/juliacon2024/talk/V8VGK9/
END:VEVENT
BEGIN:VEVENT
SUMMARY:RunwayPNPSolve.jl: Uncertainty-Aware Pose Estimation. - Romeo Vale
 ntin
DTSTART;TZID=Europe/Amsterdam:20240710T171000
DTEND;TZID=Europe/Amsterdam:20240710T172000
DTSTAMP:20260809T104232Z
UID:pretalx-juliacon2024-MVFVGB@pretalx.com
DESCRIPTION:We present `RunwayPNPSolve.jl`\, a framework for uncertainty-a
 ware pose estimation for visual landing applications with multiple methods
  including real-time least-squares minimization + resampling\, Monte-Carlo
  Markov-Chain\, and a linear approximation\, by leveraging the existing Ju
 lia package ecosystem.\nThe package further provides a framework of useful
  primitives to build simultaneously differentiable\, unitful and coordinat
 e-system-aware interfaces\, and an interactive visualization pipeline.
LOCATION:While Loop (4.2)
URL:https://pretalx.com/juliacon2024/talk/MVFVGB/
END:VEVENT
BEGIN:VEVENT
SUMMARY:ThreadedDenseSparseMul.jl: Multi-threaded Dense-Sparse Matmul. - R
 omeo Valentin
DTSTART;TZID=Europe/Amsterdam:20240710T190000
DTEND;TZID=Europe/Amsterdam:20240710T193000
DTSTAMP:20260809T104232Z
UID:pretalx-juliacon2024-VJEVSJ@pretalx.com
DESCRIPTION:We present `ThreadedDenseSparseMul.jl`\, a library that effici
 ently computes dense-sparse multiplications and outperform competing packa
 ges (and `Base.SparseArrays`) in about 20 lines of code (for the basic fun
 ctionality) by leveraging `Polyester.jl`. We further discuss the effect of
  Julia's memory layout on the performance and analyze the influence of dif
 ferent threading models.
LOCATION:Method (1.5)
URL:https://pretalx.com/juliacon2024/talk/VJEVSJ/
END:VEVENT
END:VCALENDAR