Neural DNF-MT: A neuro-symbolic approach for learning interpretable and editable policies
File(s)
Published version
Author(s)
Baugh, Kexin
Dickens, Luke
Russo, Alessandra
Type
Conference Paper
Abstract
Although deep reinforcement learning has been shown to be effective, the model’s black-box nature presents barriers to direct policy interpretation. To address this problem, we propose a neuro-symbolic approach called neural DNF-MT for end-to-end policy learning. The differentiable nature of the neural DNF-MT model enables the use of deep actor-critic algorithms for training. At the same time, its architecture is designed so that trained models can
be directly translated into interpretable policies expressed as standard (bivalent or probabilistic) logic programs. Moreover, additional layers can be included to extract abstract features from complex observations, acting as a form of predicate invention. The logic representations are highly interpretable, and we show how the bivalent
representations of deterministic policies can be edited and incorporated back into a neural model, facilitating manual intervention and adaptation of learned policies. We evaluate our approach on a range of tasks requiring learning deterministic or stochastic behaviours from various forms of observations. Our empirical results show that our neural DNF-MT model performs at the level of competing
black-box methods whilst providing interpretable policies.
be directly translated into interpretable policies expressed as standard (bivalent or probabilistic) logic programs. Moreover, additional layers can be included to extract abstract features from complex observations, acting as a form of predicate invention. The logic representations are highly interpretable, and we show how the bivalent
representations of deterministic policies can be edited and incorporated back into a neural model, facilitating manual intervention and adaptation of learned policies. We evaluate our approach on a range of tasks requiring learning deterministic or stochastic behaviours from various forms of observations. Our empirical results show that our neural DNF-MT model performs at the level of competing
black-box methods whilst providing interpretable policies.
Date Issued
2025-05-01
Date Acceptance
2024-12-19
Citation
24th International Conference on Autonomous Agents and Multiagent Systems, 2025, pp.252-260
ISBN
9798400714269
Publisher
International Foundation for Autonomous Agents and Multiagent Systems (IFAAMAS)
Start Page
252
End Page
260
Journal / Book Title
24th International Conference on Autonomous Agents and Multiagent Systems
Copyright Statement
©2025 International Foundation for Autonomous Agents and Multiagent Systems (www.ifaamas.org). This work is licensed under a Creative Commons Attribution Inter national 4.0 License.
License URL
Sponsor
US Army (US)
Engineering & Physical Science Research Council (E
Grant Number
0160 G LB679
EP/X040518/1
Source
International Conference on Autonomous Agents and Multiagent Systems
Place of Publication
Detroit, MI, USA
Publication Status
Published
Start Date
2025-05-19
Finish Date
2025-05-23
Coverage Spatial
Detroit, Michigan, USA
Date Publish Online
2025-06-05