Optimal Sparse Regression Trees

Regression trees are one of the oldest forms AI models, and their predictions can be made without a calculator, which makes them broadly useful, particularly for high-stakes applications. Within large literature on regression trees, there has been little effort towards full provable optimization, mainly due to computational hardness problem. This work proposes dynamic programming-with-bounds approach construction provably-optimal sparse trees. We leverage novel lower bound based an optimal solution k-Means clustering algorithm dimensional data. often able find in seconds, even challenging datasets that involve numbers samples highly-correlated features.