Co-Activation Graph Analysis of Safety-Verified and Explainable Deep Reinforcement Learning Policies

Dennis Gross, Helge Spieker

2025

Abstract

Deep reinforcement learning (RL) policies can demonstrate unsafe behaviors and are challenging to interpret. To address these challenges, we combine RL policy model checking—a technique for determining whether RL policies exhibit unsafe behaviors—with co-activation graph analysis—a method that maps neural network inner workings by analyzing neuron activation patterns—to gain insight into the safe RL policy’s sequential decision-making. This combination lets us interpret the RL policy’s inner workings for safe decision-making. We demonstrate its applicability in various experiments.

Paper Citation

in Harvard Style

Gross D. and Spieker H. (2025). Co-Activation Graph Analysis of Safety-Verified and Explainable Deep Reinforcement Learning Policies. In Proceedings of the 17th International Conference on Agents and Artificial Intelligence - Volume 2: ICAART; ISBN 978-989-758-737-5, SciTePress, pages 611-621. DOI: 10.5220/0013255500003890

in Bibtex Style

@conference{icaart25,
author={Dennis Gross and Helge Spieker},
title={Co-Activation Graph Analysis of Safety-Verified and Explainable Deep Reinforcement Learning Policies},
booktitle={Proceedings of the 17th International Conference on Agents and Artificial Intelligence - Volume 2: ICAART},
year={2025},
pages={611-621},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0013255500003890},
isbn={978-989-758-737-5},
}

in EndNote Style

TY - CONF

JO - Proceedings of the 17th International Conference on Agents and Artificial Intelligence - Volume 2: ICAART
TI - Co-Activation Graph Analysis of Safety-Verified and Explainable Deep Reinforcement Learning Policies
SN - 978-989-758-737-5
AU - Gross D.
AU - Spieker H.
PY - 2025
SP - 611
EP - 621
DO - 10.5220/0013255500003890
PB - SciTePress