Research map / Learning theory and algorithms / Prediction and decision losses
MAPTree: Beating “Optimal” Decision Trees with Bayesian Decision Trees
Colin Sullivan, Mo Tiwari, Sebastian Thrun
Read the paperCheck your idea against 137,217 papers
AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Frames MAP inference of Bayesian decision trees (BCART posterior) as AND/OR search and solves it with best-first search, yielding an anytime optimal-tree learner called MAPTree.
Problem
Learning optimal decision trees is computationally hard; existing optimal decision tree methods (e.g., DL8.5, GOSDT) rely on caching branch-and-bound and lack anytime behavior, while greedy methods lack optimality guarantees.
Approach
Connects MAP inference of decision trees under a Bayesian CART posterior to AND/OR search, enabling best-first AND/OR search with edge costs and fixed-size 128-bit hashed subproblem caching, giving anytime behavior and global optimality when run to completion.
Limitations
- Scalability not the main contribution; expected to be slower than greedy methods and comparable to efficient optimal methods
- Memory consumption is handled only by time or node-expansion limits; no garbage collection
- Only binary classification is covered by the current BCART-based objective
Experimental setup
Baselines: DL8.5 GOSDT CART OCT FlowOCT BinOCT
Builds on: DL8.5 GOSDT BCART
Closest papers by meaning
- Generative Model for Decision Trees (AAAI 2024 · similarity 0.89)
- Near-Optimal Decision Trees in a SPLIT Second (ICML 2025 · similarity 0.88)
- Branches: Efficiently Seeking Optimal Sparse Decision Trees via AO* (ICML 2025 · similarity 0.88)
- Optimal Decision Tree Pruning Revisited: Algorithms and Complexity (ICML 2025 · similarity 0.88)
- Learning Small Decision Trees with Few Outliers: A Parameterized Perspective (AAAI 2024 · similarity 0.87)
- Leveraging Predictive Equivalence in Decision Trees (ICML 2025 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
