Abstract
Today’s advanced Reinforcement Learning algorithms produce black-box policies, that are often difficult to interpret and trust for a person. We introduce a policy distilling algorithm, building on the CN2 rule mining algorithm, that distills the policy into a rule-based decision system. At the core of our approach is the fact that an RL process does not just learn a policy, a mapping from states to actions, but also produces extra meta-information, such as action values indicating the quality of alternative actions. This meta-information can indicate whether more than one action is near-optimal for a certain state. We extend CN2 to make it able to leverage knowledge about equally-good actions to distill the policy into fewer rules, increasing its interpretability by a person. Then, to ensure that the rules explain a valid, non-degenerate policy, we introduce a refinement algorithm that fine-tunes the rules to obtain good performance when executed in the environment. We demonstrate the applicability of our algorithm on the Mario AI benchmark, a complex task that requires modern reinforcement learning algorithms including neural networks. The explanations we produce capture the learned policy in only a few rules, that allow a person to understand what the black-box agent learned. Source code: https://gitlab.ai.vub.ac.be/yocoppen/svcn2.
Original language | English |
---|---|
Title of host publication | Trustworthy AI – Integrating Learning, Optimization and Reasoning - First International Workshop, TAILOR 2020, Revised Selected Papers |
Editors | Fredrik Heintz, Michela Milano, Barry O’Sullivan |
Publisher | Springer |
Pages | 163-179 |
Number of pages | 17 |
ISBN (Print) | 9783030739584 |
DOIs | |
Publication status | Published - 2021 |
Event | 1st International Workshop on Trustworthy AI – Integrating Learning, Optimization and Reasoning, TAILOR 2020 held as a part of European Conference on Artificial Intelligence, ECAI 2020 - Virtual, Online Duration: 4 Sept 2020 → 5 Sept 2020 |
Publication series
Name | Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) |
---|---|
Volume | 12641 LNAI |
ISSN (Print) | 0302-9743 |
ISSN (Electronic) | 1611-3349 |
Conference
Conference | 1st International Workshop on Trustworthy AI – Integrating Learning, Optimization and Reasoning, TAILOR 2020 held as a part of European Conference on Artificial Intelligence, ECAI 2020 |
---|---|
City | Virtual, Online |
Period | 4/09/20 → 5/09/20 |
Bibliographical note
Green Open Access added to TU Delft Institutional Repository ‘You share, we take care!’ – Taverne project https://www.openaccess.nl/en/you-share-we-take-careOtherwise as indicated in the copyright section: the publisher is the copyright holder of this work and the author uses the Dutch legislation to make this work public.
Keywords
- Explainable AI
- Inductive rule learning
- Policy distillation
- Reinforcement Learning