- add more tests
- implement gradient boost and adaboost for regression
- test weights support for both regression and classification trees
- add min_weights_leaf prepruning
- add stricter typing for leaf and node api
- develop "compact" leaf, which only holds counts of classes
- add new options for user input purity criterions
- optimize vectorization, e.g, changing
nc[:] .= 0to loops - add postpruning option with comparison with validation data
- use range [0,1] for purity_thresh in new implementations of
prune_tree(currently commented out) - standardize variable names to snake case
- trees should still split if purity change is equal to min_purity_increase
- add benchmarks for other functions