Publikation
EXPIL: Explanatory Predicate Invention for Learning in Games
Jingyuan Sha; Hikaru Shindo; Quentin Delfosse; Kristian Kersting; Devendra Singh Dhami
In: Computing Research Repository eprint Journal (CoRR), Vol. abs/2406.06107, Pages 1-11, arXiv, 2024.
Zusammenfassung
Reinforcement learning (RL) has proven to be a powerful tool
for training agents that excel in various games. However, the
black-box nature of neural network models often hinders our
ability to understand the reasoning behind the agent’s actions.
Recent research has attempted to address this issue by us-
ing the guidance of pretrained neural agents to encode logic-
based policies, allowing for interpretable decisions. A draw-
back of such approaches is the requirement of large amounts
of predefined background knowledge in the form of predi-
cates, limiting its applicability and scalability. In this work,
we propose a novel approach, Explanatory Predicate Inven-
tion for Learning in Games (EXPIL), that identifies and ex-
tracts predicates from a pretrained neural agent, later used in
the logic-based agents, reducing the dependency on prede-
fined background knowledge. Our experimental evaluation
on various games demonstrate the effectiveness of EXPIL in
achieving explainable behavior in logic agents while requir-
ing less background knowledge.