Skip to content

mmschlk/TreeSHAP-IQ

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

⭐Supplement Implementation

Paper: Beyond TreeSHAP: Efficient Computation of any-order Shapley Interactions for Tree Ensembles


🔧Installation

The implementation was written and tested with python version 3.9.7. Install the dependencies via pip:

pip install -r requirements.txt

📊Running the experiments

The main experiment functionality is implemented in experiments_main.py including all plots. The scripts for running the experiments on the individual datasets are named exp_<dataset>.py. Therein, you find the code for pre-processing, training, and explaining.

🔨Errors in outdated SHAP

Warning: If you want to compare the results with the original TreeSHAP implementation, you need to install the original TreeSHAP implementation from pip and change a couple of lines of code in there.

Change line 250 in _tree.py from

X_missing = np.isnan(X, dtype=np.bool)

to

X_missing = np.isnan(X, dtype=bool)

and change line 1102 in _tree.py from

X_missing = np.isnan(X, dtype=np.bool)

to

X_missing = np.isnan(X, dtype=bool)

and change line 82 in _tabular.py from

self._last_mask = np.zeros(data.shape[1], dtype=np.bool)

to

self._last_mask = np.zeros(data.shape[1], dtype=bool)

About

Supplement Material for research project

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages