A0636
Title: Geometric shapes of the tree-induced partition
Authors: Hengrui Luo - Rice University (United States) [presenting]
Abstract: The purpose is to start by explaining the geometric nature of these partitions, illustrating how decision trees essentially draw multidimensional boundaries to segregate them by partitioning the input space. Expanding on this, the setting of tensor-based inputs and outputs is moved into, showcasing the complex geometric shapes that emerge and the new challenges they bring to the tree-based models. To navigate these complexities, the innovative tensor-on-tensor tree regression approach is introduced, designed to adeptly manage this multidimensional geometric partitioning. In conclusion, a step is taken back to reflect on the fundamental geometric principles underpinning tree-induced partitions and ponder future research avenues in this fascinating intersection of geometry and tree-based models.
