A great many thanks to the BourbakIA winning team (Jean-Matthieu Schertzer, Cédric Faucheux, Jean Cupe, Nicolas Béguin and David Cortés) ! this methodology and prototype which has led to AntakIA is the direct following of that 2-day hackathon organised by the French financial regulator (ACPR). Special thanks to Jean-Matthieu, for his great work developping Skope-Rules...
We are also very grateful to the Confiance.AI national French initiative, for helping us try this approach with great French industrial actors, to increase adoption and robustness of high risk models.
We also try the calculation of Banzhaf indices (not only Shapley) more stable and less computer intensive following our ongoing discussion with the CRIL (Centre de Recherche en Informatique de Lens, Pierre Marquis, Gilles Audemard, Jean-Marie Lagniez, Steve Bellart, Louenas Bounia). Quite a lot more to come ! and do have a look at this quite promising initiative on X-AI by the CRIL
Special thanks to the Jupyter project team. Without Jupyter, some particular AntakIA wouldn't be possible : ipywidgets or ipyvuetify from the great Widgetti team ...
Thanks to SoyHuCe, Weituo Dai, Antoine Edy, Xavier Lekeuka for paving the way !