Skip to content
DraconiticResearch
Search Sources API
201,683 studies31 sourcesUpdated 42 min ago
PreprintCC BY 4.0arXiv

Coordinate-Free Automatic Differentiation

  • Alex Guerra
  • Joshua Aduol
  • Yaqian Tang
  • Daniel Williams
  • Ryan P. Adams
  • Peter Orbanz
Open at sourcePDFDOICite

Abstract

Automatic differentiation (AD) is a cornerstone of modern machine learning. Since all AD frameworks implicitly fix a basis of Euclidean space to define gradients and Jacobians, they natively process mathematical objects that are arrays of a fixed set of coordinates. This makes it awkward to express objects such as rotations, volume forms, and curvatures that arise, for example, in applications of AI to science. Bespoke workarounds lack reusability. We show that (i) a single, basis-independent derivative operator can represent many geometric differential operations, and (ii) this operator can be expressed entirely in terms of existing AD primitives. This reusable transformation can be composed with JAX batching, compilation, and hardware acceleration. Geometry enters as algebraic data. We experimentally demonstrate its reuse across different manifold operators, its composition through a spacetime inverse-problem pipeline, and its use in optimization over different rotor subgroups.

Related

  1. Derivative Computation in PINNs: Automatic Differentiation, Finite Differences and Beyond
    Maciej J. Mikulski, Tadeusz Uhl
    arXivPreprintarXiv licenceFull textNumerical analysis & scientific computingComputational physicsarXiv
  2. Automatic Differentiation with Constraints in Forward and Inverse Problems
    Nikolay Yavich, Alexander Ryabov, Viacheslav Naumov +4
    arXivPreprintCC BY-NC-ND 4.0Full textNumerical analysis & scientific computingarXiv
  3. asdex: Automatic Sparse Differentiation in JAX
    Adrian Hill, Guillaume Dalle
    arXivPreprintCC BY 4.0Full textNumerical analysis & scientific computingarXiv
  4. Shape calculus and automatic differentiation for multi-phase level-set topology optimisation with unfitted finite elements
    Zachary J. Wegert, Martin Berggren, Vivien J. Challis
    arXivPreprintCC BY 4.0Full textNumerical analysis & scientific computingOptimisation & controlarXiv
  5. Freely Generated Categorical Structures and Automatic Differentiation, PhD Thesis (Introduction and Conclusion)
    Fernando Lucatelli Nunes
    arXivPreprintarXiv licenceFull textAlgebra & number theoryarXiv

Cite

BibTeXRIS
BibTeX
@misc{guerra2026coordinatefree,
  title = {{Coordinate-Free Automatic Differentiation}},
  author = {Alex Guerra and Joshua Aduol and Yaqian Tang and Daniel Williams and Ryan P. Adams and Peter Orbanz},
  howpublished = {arXiv},
  year = {2026},
  month = oct,
  doi = {10.48550/arxiv.2610.12776},
  eprint = {2610.12776},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2610.12776},
  note = {CC BY 4.0},
}
RIS
TY  - UNPB
TI  - Coordinate-Free Automatic Differentiation
AU  - Alex Guerra
AU  - Joshua Aduol
AU  - Yaqian Tang
AU  - Daniel Williams
AU  - Ryan P. Adams
AU  - Peter Orbanz
PY  - 2026
DA  - 2026/10/08
PB  - arXiv
DO  - 10.48550/arxiv.2610.12776
UR  - https://arxiv.org/abs/2610.12776
AB  - Automatic differentiation (AD) is a cornerstone of modern machine learning. Since all AD frameworks implicitly fix a basis of Euclidean space to define gradients and Jacobians, they natively process mathematical objects that are arrays of a fixed set of coordinates. This makes it awkward to express objects such as rotations, volume forms, and curvatures that arise, for example, in applications of AI to science. Bespoke workarounds lack reusability. We show that (i) a single, basis-independent derivative operator can represent many geometric differential operations, and (ii) this operator can be expressed entirely in terms of existing AD primitives. This reusable transformation can be composed with JAX batching, compilation, and hardware acceleration. Geometry enters as algebraic data. We experimentally demonstrate its reuse across different manifold operators, its composition through a spacetime inverse-problem pipeline, and its use in optimization over different rotor subgroups.
ER  -