Interpretable and Editable Programmatic Tree Policies for Reinforcement Learning

Hector Kohler, Quentin Delfosse, Riad Akrour, Kristian Kersting, Philippe Preux
5/23/2024
cs.AIcs.LG

No official code found. Best implementation: KohlerHECTOR/interpreter-py (15 stars, MIT, last commit 2024-09-10).

Abstract

Deep reinforcement learning agents are prone to goal misalignments. The black-box nature of their policies hinders the detection and correction of such misalignments, and the trust necessary for real-world deployment. So far, solutions learning interpretable policies are inefficient or require many human priors. We propose INTERPRETER, a fast distillation method producing INTerpretable Editable tRee Programs for ReinforcEmenT lEaRning. We empirically demonstrate that INTERPRETER compact tree programs match oracles across a diverse set of sequential decision tasks and evaluate the impact of our design choices on interpretability and performances. We show that our policies can be interpreted and edited to correct misalignments on Atari games and to explain real farming strategies.

View on arXivView PDF

Code Implementations(1)

KohlerHECTOR/interpreter-pyCommunity100%
MIT

Implementation of Interpretable and Editable Programmatic Tree Policies for Reinforcement Learning (Kohler, Delfosse, et. al. 2024).

151PythonJul 3, 20242 years agoMIT
code-generationexplainabilityexplainable-aiimitation-learninginterpretability+4 more

Similar Papers

Towards Automated Semantic Interpretability in Reinforcement Learning via Vision-Language Models

85% match2025

Evaluating Interpretable Reinforcement Learning by Distilling Policies into Programs

84% match2025

Synthesizing Programmatic Policies with Actor-Critic Algorithms and ReLU Networks

83% match2023

Optimizing Interpretable Decision Tree Policies for Reinforcement Learning

83% match1 repo102024

BET: Explaining Deep Reinforcement Learning through The Error-Prone Decisions

83% match2024

"So, Tell Me About Your Policy...": Distillation of interpretable policies from Deep Reinforcement Learning agents

83% match2025

Cite this paper

@article{kohler2024interpretable,
  title  = {Interpretable and Editable Programmatic Tree Policies for Reinforcement Learning},
  author = {Hector Kohler and Quentin Delfosse and Riad Akrour and Kristian Kersting and Philippe Preux},
  year   = {2024},
  eprint = {2405.14956},
  archivePrefix = {arXiv},
  url    = {http://arxiv.org/abs/2405.14956v1}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2405.14956.svg)](https://codeofpaper.com/paper/2405.14956)

Discussion