/Sparti

Sparti - Spatial Partition Inference Toolbox - is a statistical software package for spatial partition inference such as inference methods for the Mondrian Process, the Binary Space Partitioning-Tree Process, and the Rectangular Bounding Process.

Primary LanguagePython

Sparti - Spatial Partition Inference Toolbox

A toolbox of Bayesian nonparametric space partition models. We provide implementations of Sparti methods as follows:

  • Rectangular Bounding Process Xuhui Fan, Bin Li, Scott A. Sisson. The 32nd Conference on Neural Information Processing Systems (NeurIPS-18)

    This method uses "bounding" strategy to directly generate boxes in a product space and is able to reduce modelling on sparse and noisy regions.

  • Binary Space Partitioning-Tree Process Xuhui Fan, Bin Li, Scott A. Sisson. The 21st International Conference on Artificial Intelligence and Statistics (AISTATS- 18).

    This method allows sloped hierarchical cuts in a 2-dimensional space and can describe dimensional dependencies.

  • Binary Space Partitioning Forest Xuhui Fan, Bin Li, Scott A. Sisson. The 22nd International Conference on Artificial Intelligence and Statistics (AISTATS- 19).

    This method uses 2-dimensional hyperplane to enable sloped cuts in a multi-dimensional space.

  • Online Binary Space Partitioning Forests Xuhui Fan, Bin Li, Scott A. Sisson. The 23rd International Conference on Artificial Intelligence and Statistics (AISTATS-20).

    This method adopts the generative process of Binary Space Partitioning Forests to enable online learning.