Tractable Explanations for d-DNNF Classifiers - Archive ouverte HAL Accéder directement au contenu
Communication Dans Un Congrès Année : 2022

Tractable Explanations for d-DNNF Classifiers

Résumé

Compilation into propositional languages finds a growing number of practical uses, including in constraint programming, diagnosis and machine learning (ML), among others. One concrete example is the use of propositional languages as classifiers, and one natural question is how to explain the predictions made. This paper shows that for classifiers represented with some of the best-known propositional languages, different kinds of explanations can be computed in polynomial time. These languages include deterministic decomposable negation normal form (d-DNNF), and so any propositional language that is strictly less succinct than d-DNNF. Furthermore, the paper describes optimizations, specific to Sentential Decision Diagrams (SDDs), which are shown to yield more efficient algorithms in practice.
Fichier principal
Vignette du fichier
20514-Article Text-24527-1-2-20220628.pdf (150.92 Ko) Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)

Dates et versions

hal-03873826 , version 1 (27-11-2022)

Identifiants

Citer

Xuanxiang Huang, Yacine Izza, Alexey Ignatiev, Martin Cooper, Nicholas Asher, et al.. Tractable Explanations for d-DNNF Classifiers. 36th AAAI Conference on Artificial Intelligence (AAAI 2022), AAAI: American Association for Artificial Intelligence, Feb 2022, Seattle (virtual), United States. pp.5719-5728, ⟨10.1609/aaai.v36i5.20514⟩. ⟨hal-03873826⟩
39 Consultations
46 Téléchargements

Altmetric

Partager

Gmail Facebook X LinkedIn More