Automatic Differentiation via Effects and Handlers: An Implementation in Frank
Programming Languages
2021-01-21 v1 Machine Learning
Abstract
Automatic differentiation (AD) is an important family of algorithms which enables derivative based optimization. We show that AD can be simply implemented with effects and handlers by doing so in the Frank language. By considering how our implementation behaves in Frank's operational semantics, we show how our code performs the dynamic creation of programs during evaluation.
Cite
@article{arxiv.2101.08095,
title = {Automatic Differentiation via Effects and Handlers: An Implementation in Frank},
author = {Jesse Sigal},
journal= {arXiv preprint arXiv:2101.08095},
year = {2021}
}
Comments
Appeared as short paper in PEPM'21, see https://www.youtube.com/watch?v=BmBSJFkfL2M for associated talk