Simple, Accurate, and Efficient Axis-Aligned Decision Tree Learning
Keywords: decision tree, gradient descent, tabular data
Abstract: Decision Trees (DTs) are widely used in various domains for their simplicity and interpretability. However, traditional DTs often suffer from low accuracy and reduced robustness because they rely on fixed splits and a greedy approach to decision-making. While recent approaches combining decision trees with optimization seek to balance accuracy, computational efficiency, and interpretability, they still fall short. In this paper, we introduce a novel Probabilistic univariate Decision Tree (ProuDT), a non-greedy, axis-aligned tree that aims to address these challenges and achieve significant improvements. By assigning a single deterministic feature to each decision node, ProuDT ensures univariate splits while preserving the differentiability of soft decision trees for gradient-based optimization. This tree enhances interpretability through transparent feature utilization in decision-making. Additionally, ProuDT simplifies the optimization process and reduces computational cost by avoiding complex parameters. Extensive experiments on tabular datasets demonstrate ProuDT’s superior performance and scalability in binary and multi-class classification tasks.
Primary Area: other topics in machine learning (i.e., none of the above)
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.
Reciprocal Reviewing: I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6.
Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.
No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.
Submission Number: 13015
Loading