A Common Interface for Automatic Differentiation

Guillaume Dalle, Adrian Hill
5/8/2025
cs.MScs.LGmath.NA

No official code found. Best implementation: JuliaDiff/DifferentiationInterface.jl (310 stars, MIT, last commit 2026-04-29).

Abstract

For scientific machine learning tasks with a lot of custom code, picking the right Automatic Differentiation (AD) system matters. Our Julia package DifferentiationInterface$.$jl provides a common frontend to a dozen AD backends, unlocking easy comparison and modular development. In particular, its built-in preparation mechanism leverages the strengths of each backend by amortizing one-time computations. This is key to enabling sophisticated features like sparsity handling without putting additional burdens on the user.

View on arXivView PDF

Code Implementations(1)

JuliaDiff/DifferentiationInterface.jlCommunity100%
MIT

An interface to various automatic differentiation backends in Julia.

31033JuliaJan 9, 20245 months agoMIT
autodiffautomatic-differentiationdifferentiationjuliamachine-learning

Similar Papers

DaCe AD: Unifying High-Performance Automatic Differentiation for Machine Learning and Scientific Computing

83% match2025

Forward-Mode Automatic Differentiation in Julia

79% match1 repo1.0k2016

A Step-by-step Introduction to the Implementation of Automatic Differentiation

78% match2024

A Benchmark of Selected Algorithmic Differentiation Tools on Some Problems in Computer Vision and Machine Learning

76% match1 repo32018

TextGrad: Automatic "Differentiation" via Text

75% match1 repo3.7k2024

Marian: Fast Neural Machine Translation in C++

74% match1 repo1.5k2018

Cite this paper

@article{dalle2025common,
  title  = {A Common Interface for Automatic Differentiation},
  author = {Guillaume Dalle and Adrian Hill},
  year   = {2025},
  eprint = {2505.05542},
  archivePrefix = {arXiv},
  url    = {http://arxiv.org/abs/2505.05542v1}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2505.05542.svg)](https://codeofpaper.com/paper/2505.05542)

Discussion