Formal Explanations of Neural Network Policies for Planning - LAAS - Laboratoire d'Analyse et d'Architecture des Systèmes Accéder directement au contenu
Communication Dans Un Congrès Année : 2023

Formal Explanations of Neural Network Policies for Planning

Résumé

Deep learning is increasingly used to learn policies for planning problems, yet policies represented by neural networks are difficult to interpret, verify and trust. Existing formal approaches to post-hoc explanations provide concise reasons for a single decision made by an ML model. However, understanding planning policies require explaining sequences of decisions. In this paper, we formulate the problem of finding explanations for the sequence of decisions recommended by a learnt policy in a given state. We show that, under certain assumptions, a minimal explanation for a sequence can be computed by solving a number of single decision explanation problems which is linear in the length of the sequence. We present experimental results of our implementation of this approach for ASNet policies for classical planning domains.
Fichier principal
Vignette du fichier
ijcai23.pdf (388.28 Ko) Télécharger le fichier
Origine Fichiers produits par l'(les) auteur(s)

Dates et versions

hal-04241579 , version 1 (13-10-2023)

Licence

Copyright (Tous droits réservés)

Identifiants

Citer

Renee Selvey, Alban Grastien, Sylvie Thiébaux. Formal Explanations of Neural Network Policies for Planning. Thirty-Second International Joint Conference on Artificial Intelligence {IJCAI-23}, Aug 2023, Macau, Macau SAR China. pp.5446-5456, ⟨10.24963/ijcai.2023/605⟩. ⟨hal-04241579⟩
97 Consultations
85 Téléchargements

Altmetric

Partager

Gmail Mastodon Facebook X LinkedIn More