Deduction with Induction: Combining Knowledge Discovery and Reasoning for Interpretable Deep Reinforcement Learning

Abstract

Deep reinforcement learning (DRL) has achieved remarkable success in dynamic decision-making tasks. However, its inherent opacity and cold start problem hinder transparency and training efficiency. To address these challenges, we propose HRL-ID, a neural-symbolic framework that combines automated rule discovery with logical reasoning within a hierarchical DRL structure. HRL-ID dynamically extracts first-order logic rules from environmental interactions, iteratively refines them through success-based updates, and leverages these rules to guide action execution during training. Extensive experiments on Atari benchmarks demonstrate that HRL-ID outperforms state-of-the-art methods in training efficiency and interpretability, achieving higher reward rates and successful knowledge transfer between domains.

Cite

Text

Zhang et al. "Deduction with Induction: Combining Knowledge Discovery and Reasoning for Interpretable Deep Reinforcement Learning." International Joint Conference on Artificial Intelligence, 2025. doi:10.24963/IJCAI.2025/777

Markdown

[Zhang et al. "Deduction with Induction: Combining Knowledge Discovery and Reasoning for Interpretable Deep Reinforcement Learning." International Joint Conference on Artificial Intelligence, 2025.](https://mlanthology.org/ijcai/2025/zhang2025ijcai-deduction/) doi:10.24963/IJCAI.2025/777

BibTeX

@inproceedings{zhang2025ijcai-deduction,
  title     = {{Deduction with Induction: Combining Knowledge Discovery and Reasoning for Interpretable Deep Reinforcement Learning}},
  author    = {Zhang, Haodi and Zeng, Xiangyu and Chen, Junyang and Song, Yuanfeng and Mao, Rui and Lin, Fangzhen},
  booktitle = {International Joint Conference on Artificial Intelligence},
  year      = {2025},
  pages     = {6984-6992},
  doi       = {10.24963/IJCAI.2025/777},
  url       = {https://mlanthology.org/ijcai/2025/zhang2025ijcai-deduction/}
}