/dDBCSFi2

Code of "Shap Scores for Neural Network Classifiers: Black-Box vs. Open-Box Circuit Compilation"

Primary LanguageJupyter Notebook

dDBCSFi(2)

Code of "Opening Up the Neural Network Classifier for Shap Score Computation"/"Compiling Neural Network Classifiers into Boolean Circuits for Efficient Shap-Score Computation", written by Jorge E. León, with the guidance of Leopoldo Bertossi. Feel free to use it as you please 👍

This code is thought to be used on Google Colab. Modifications should be made to be installed on other computers (especially the ones with an OS different from Linux).

A big thanks goes to Arthur Choi, Norbert Manthey & Andy Shih for sharing their respective repositories, which were key to the development of this code. These can be found down below:

Also thanks to Maximilian Schleich for his non public code that he kindly shared with us.