Generalized Integrated Gradients: A practical method for explaining diverse ensembles
This provides a practical explanation method for mixed-type models in domains like financial services, though it appears incremental as an extension of existing techniques.
The authors tackled the problem of explaining diverse ensembles in machine learning by introducing Generalized Integrated Gradients (GIG), an extension of Integrated Gradients, and proved it as the only correct method under specific axioms for mixed-type models.
We introduce Generalized Integrated Gradients (GIG), a formal extension of the Integrated Gradients (IG) (Sundararajan et al., 2017) method for attributing credit to the input variables of a predictive model. GIG improves IG by explaining a broader variety of functions that arise from practical applications of ML in domains like financial services. GIG is constructed to overcome limitations of Shapley (1953) and Aumann-Shapley (1974), and has desirable properties when compared to other approaches. We prove GIG is the only correct method, under a small set of reasonable axioms, for providing explanations for mixed-type models or games. We describe the implementation, and present results of experiments on several datasets and systems of models.