Optimization Methods for Interpretable Differentiable Decision Trees Applied to Reinforcement Learning

Andrew Silva, Matthew Gombolay, Taylor Killian, Ivan Jimenez, Sung-Hyun Son
Proceedings of the Twenty Third International Conference on Artificial Intelligence and Statistics, PMLR 108:1855-1865, 2020.

Abstract

Decision trees are ubiquitous in machine learning for their ease of use and interpretability. Yet, these models are not typically employed in reinforcement learning as they cannot be updated online via stochastic gradient descent. We overcome this limitation by allowing for a gradient update over the entire tree that improves sample complexity affords interpretable policy extraction. First, we include theoretical motivation on the need for policy-gradient learning by examining the properties of gradient descent over differentiable decision trees. Second, we demonstrate that our approach equals or outperforms a neural network on all domains and can learn discrete decision trees online with average rewards up to 7x higher than a batch-trained decision tree. Third, we conduct a user study to quantify the interpretability of a decision tree, rule list, and a neural network with statistically significant results (p < 0.001).

Cite this Paper


BibTeX
@InProceedings{pmlr-v108-silva20a, title = {Optimization Methods for Interpretable Differentiable Decision Trees Applied to Reinforcement Learning}, author = {Silva, Andrew and Gombolay, Matthew and Killian, Taylor and Jimenez, Ivan and Son, Sung-Hyun}, booktitle = {Proceedings of the Twenty Third International Conference on Artificial Intelligence and Statistics}, pages = {1855--1865}, year = {2020}, editor = {Chiappa, Silvia and Calandra, Roberto}, volume = {108}, series = {Proceedings of Machine Learning Research}, month = {26--28 Aug}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v108/silva20a/silva20a.pdf}, url = {https://proceedings.mlr.press/v108/silva20a.html}, abstract = { Decision trees are ubiquitous in machine learning for their ease of use and interpretability. Yet, these models are not typically employed in reinforcement learning as they cannot be updated online via stochastic gradient descent. We overcome this limitation by allowing for a gradient update over the entire tree that improves sample complexity affords interpretable policy extraction. First, we include theoretical motivation on the need for policy-gradient learning by examining the properties of gradient descent over differentiable decision trees. Second, we demonstrate that our approach equals or outperforms a neural network on all domains and can learn discrete decision trees online with average rewards up to 7x higher than a batch-trained decision tree. Third, we conduct a user study to quantify the interpretability of a decision tree, rule list, and a neural network with statistically significant results (p < 0.001).} }
Endnote
%0 Conference Paper %T Optimization Methods for Interpretable Differentiable Decision Trees Applied to Reinforcement Learning %A Andrew Silva %A Matthew Gombolay %A Taylor Killian %A Ivan Jimenez %A Sung-Hyun Son %B Proceedings of the Twenty Third International Conference on Artificial Intelligence and Statistics %C Proceedings of Machine Learning Research %D 2020 %E Silvia Chiappa %E Roberto Calandra %F pmlr-v108-silva20a %I PMLR %P 1855--1865 %U https://proceedings.mlr.press/v108/silva20a.html %V 108 %X Decision trees are ubiquitous in machine learning for their ease of use and interpretability. Yet, these models are not typically employed in reinforcement learning as they cannot be updated online via stochastic gradient descent. We overcome this limitation by allowing for a gradient update over the entire tree that improves sample complexity affords interpretable policy extraction. First, we include theoretical motivation on the need for policy-gradient learning by examining the properties of gradient descent over differentiable decision trees. Second, we demonstrate that our approach equals or outperforms a neural network on all domains and can learn discrete decision trees online with average rewards up to 7x higher than a batch-trained decision tree. Third, we conduct a user study to quantify the interpretability of a decision tree, rule list, and a neural network with statistically significant results (p < 0.001).
APA
Silva, A., Gombolay, M., Killian, T., Jimenez, I. & Son, S.. (2020). Optimization Methods for Interpretable Differentiable Decision Trees Applied to Reinforcement Learning. Proceedings of the Twenty Third International Conference on Artificial Intelligence and Statistics, in Proceedings of Machine Learning Research 108:1855-1865 Available from https://proceedings.mlr.press/v108/silva20a.html.

Related Material