POPL 2020 (series) / LAFI 2020 (series) / LAFI 2020: Languages for Inference (formerly PPS) /
A Differential-form Pullback Programming Language for Higher-order Reverse-mode Automatic Differentiation
Building on the observation that reverse-mode automatic differentiation (AD), a generalisation of backpropagation, can naturally be expressed as pullbacks of differential 1-forms, we design a higher-order, call-by-value functional programming language with a first-class reverse-mode AD operator. Using the category of convenient vector spaces and smooth maps, we show that the operational semantics precisely captures reverse-mode AD, even in a higher-order setting. We exhibit the connection with the differential λ-calculus via a translation that preserves reduction and interpretation.
Tue 21 JanDisplayed time zone: Saskatchewan, Central America change
Tue 21 Jan
Displayed time zone: Saskatchewan, Central America change
10:30 - 12:30 | |||
10:30 30mTalk | A Differential-form Pullback Programming Language for Higher-order Reverse-mode Automatic Differentiation LAFI | ||
11:00 30mTalk | A Monad for Point Processes LAFI File Attached | ||
11:30 30mTalk | Denotational Semantics for Differentiable Programming with Manifolds LAFI Jesse Sigal University of Edinburgh | ||
12:00 30mTalk | Backpropagation in the Simply Typed Lambda-calculus with Linear Negation LAFI |