A Geometric Theory of Higher-Order Automatic Differentiation
read the original abstract
First-order automatic differentiation is a ubiquitous tool across statistics, machine learning, and computer science. Higher-order implementations of automatic differentiation, however, have yet to realize the same utility. In this paper I derive a comprehensive, differential geometric treatment of automatic differentiation that naturally identifies the higher-order differential operators amenable to automatic differentiation as well as explicit procedures that provide a scaffolding for high-performance implementations.
This paper has not been read by Pith yet.
Forward citations
Cited by 2 Pith papers
-
Archimedean Copula Inference via Taylor-Mode AD
acopula enables polynomial-time exact inference for arbitrary nested Archimedean copulas with censoring via Taylor-mode AD on user-defined generators.
-
Higher Order Automatic Differentiation of Higher Order Functions
The paper characterizes forward-mode AD as a unique structure-preserving macro on a higher-order language with ADTs and proves its semantic correctness using a gluing construction on diffeological spaces, extending to...
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.