Parallelism-preserving automatic differentiation for second-order array languages
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 AugDisplayed time zone: Seoul change
23:30 - 01:00
|Parallelism-preserving automatic differentiation for second-order array languages|
|Reverse Automatic Differentiation for Accelerate (Extended Abstract)|
|Computing Persistent Homology in Futhark|
Erik von Brömssen Chalmers University of Technology