A Python implementation of the R Setar Tree models from Godahewa et al.

Objectives

• porting the default Setar-Trees and Setar-Forest to Python code
  • Writing Python code + tests when critical
  • Defining a sensical API
  • Writing tests + documentation
  • Accelerating the code with Numba
  • benchmarking the models
• Adding extensions to the model
  • Extra-Setar
  • Boosted-Setar?
  • Probabilistic-Setar?

Learnings

• R is frustrating to work with:
  • Lists are 1-indexed
  • Array[:, -1] means all columns except the first one ???
  • XtX = list(); XtX.left = matrix(0,p,p) is possible, like... why not?
  • True is T and False is F, like why would you want to have explicit naming in your code anyway?
  • length(1) = 1 ...