English

Adjoint Differentiation for generic matrix functions

Computational Finance 2021-09-13 v1 Symbolic Computation Risk Management

Abstract

We derive a formula for the adjoint A\overline{A} of a square-matrix operation of the form C=f(A)C=f(A), where ff is holomorphic in the neighborhood of each eigenvalue. We then apply the formula to derive closed-form expressions in particular cases of interest such as the case when we have a spectral decomposition A=UDU1A=UDU^{-1}, the spectrum cut-off C=A+C=A_+ and the Nearest Correlation Matrix routine. Finally, we explain how to simplify the computation of adjoints for regularized linear regression coefficients.

Cite

@article{arxiv.2109.04913,
  title  = {Adjoint Differentiation for generic matrix functions},
  author = {Andrei Goloubentsev and Dmitri Goloubentsev and Evgeny Lakshtanov},
  journal= {arXiv preprint arXiv:2109.04913},
  year   = {2021}
}
R2 v1 2026-06-24T05:51:45.233Z