Jacobus van der Linden
Publication highlights
Piecewise Constant and Linear Regression Trees: An Optimal Dynamic Programming Approach
Published in Proceedings of ICML-24, 2024
We use DP to generate optimal regression trees with constant and (simple) linear regression models in the leaf node. Our method improves scalability by one or more orders of magnitude in comparison to the state-of-the-art.
Recommended citation: Van den Bos, M., van der Linden, J. G. M., & Demirović, E. (2024). "Piecewise Constant and Linear Regression Trees: An Optimal Dynamic Programming Approach." Proceedings of ICML-24.
Download Paper | Source Repository
Necessary and Sufficient Conditions for Optimal Decision Trees Using Dynamic Programming
Published in Advances in NeurIPS-23, 2023
We prove necessary and sufficient conditions for the use of DP for optimal decision trees and provide a framework STreeD that can optimize trees for a variety of objectives and constraints.
Recommended citation: Van der Linden, J. G. M., de Weerdt, M. M., & Demirović, E. (2023). "Necessary and Sufficient Conditions for Optimal Decision Trees Using Dynamic Programming." Advances in NeurIPS-23, 9173-9212.
Download Paper | Source Repository