ICFP 2021
Sun 22 - Sat 28 August 2021
Sun 22 Aug 2021 23:30 - 00:00 at FHPNC - Session 2

We develop automatic differentiation (AD) procedures for reductions and scans—parameterized by arbitrary differentiable monoids—in a way that preserves parallelism, by rewriting them as other reductions and scans. This is in contrast with the literature and with existing AD systems, which are either general, but force sequential execution of the derivative program, or only include hand-crafted rules for a select few monoids (usually $(0, +)$, $(1, \times)$, $(-\infty, \max)$ and $(\infty, \min)$) and thus lack the general flexibility of second-order languages.

Sun 22 Aug

Displayed time zone: Seoul change

23:30 - 01:00
Session 2FHPNC at FHPNC
23:30
30m
Talk
Parallelism-preserving automatic differentiation for second-order array languages
FHPNC
Adam Paszke Google Research, Matthew Johnson Google Research, Roy Frostig Google Research, Dougal Maclaurin Google Research
00:00
30m
Talk
Reverse Automatic Differentiation for Accelerate (Extended Abstract)
FHPNC
Tom Smeding Utrecht University, Matthijs Vákár Utrecht University, Trevor L. McDonell Utrecht University
00:30
30m
Talk
Computing Persistent Homology in Futhark
FHPNC
Erik von Brömssen Chalmers University of Technology