Explainable Reinforcement Learning via Reward Decomposition

By Zoe Juozapaitis, Anurag Koul, Alan Fern, Martin Erwig, Finale Doshi-Velez. 2019

Subfields: Reinforcement Learning, Safety, Robustness & Interpretability

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Reinforcement Learning

1. Erika Puiutta
2. Alexandre Heuillet
3. Fabien Couthouis
4. Natalia Díaz Rodríguez
5. Zhi Chen
6. Eric M. S. P. Veith
7. Been Kim
8. Cynthia Rudin

In Safety, Robustness & Interpretability

1. Lesia Semenova
2. Zhi Chen
3. Been Kim
4. Chudi Zhong
5. Haiyang Huang
6. Natalia Díaz Rodríguez
7. Tim Miller
8. Cynthia Rudin

Citations in the corpus

• Interpretable Machine Learning: Fundamental Principles and 10 Grand Challenges (2021, methodology): “, 2018), reward decomposition (Juozapaitis et al., 2019), autonomous explanation (Fukuchi et al.”
• Explainable Reinforcement Learning Through a Causal Lens (2019, methodology): “Soft decision trees have been used to generate more interpretable policies (Coppens et al. 2019), and reward decomposition has been utilized to provide minimum sufficient explanations in RL (Juozapaitis et al. 2019).”
• Explainable Reinforcement Learning: A Survey (2020, background): “[64]; of all thirteen selected XRL methods, only five evaluate (non-expert) user satisfaction and/or utility of a method [53, 27, 64, 17, 41], and only three of these offer contrastive explanations [41, 53, 64].”
• Acquisition of chess knowledge in AlphaZero (2021, background): “Reward difference explanations [56] have shown to be helpful in understanding why certain actions are being taken instead of others.”
• Explainability in Deep Reinforcement Learning (2020, methodology): “Transparent algorithms are well known and used in standard Machine Learning (e.g., linear regression, decision trees or rule-based systems).”