Pré-Publication, Document De Travail Année : 2026

Fast and General Automatic Differentiation for Finite-State Methods

Résumé

We propose a new method, that we coined the “morphism-trick”, to integrate custom implementations of vector-Jacobian products in automatic differentiation softwares, applicable to a wide range of semiring-based computations. Our approach leads to efficient and semiring-agnostic implementations of the backward pass of dynamic programming algorithms. For the particular case of finite-state methods, we introduce an algorithm that computes and differentiates the ⊕-sum of all paths’ weight of a finite-state automaton. Results show that, with minimal effort from the user, our novel library allows computing the gradient of a function w.r.t. to the weights of a finite state automaton orders of magnitude faster than state-of-the-art automatic differentiation systems. Implementations are made available via an open-source library distributed under a permissive license.

Fichier principal
Vignette du fichier
main.pdf (499.15 Ko) Télécharger le fichier
Origine Fichiers produits par l'(les) auteur(s)
Licence

Dates et versions

hal-05500843 , version 1 (11-02-2026)

Licence

Identifiants

  • HAL Id : hal-05500843 , version 1

Citer

Lucas Ondel Yang, Tina Raissi, Martin Kocour, Pablo Riera, Caio Corro. Fast and General Automatic Differentiation for Finite-State Methods. 2026. ⟨hal-05500843⟩
520 Consultations
107 Téléchargements

Partager

  • More