A Unified Approach to Interpreting Model Predictions

Authors: Scott M. Lundberg, Su-In Lee
Year: 2017
Summary: This paper introduces SHAP (SHapley Additive exPlanations), a method to explain individual predictions. It is based on the game-theoretically optimal Shapley values. SHAP attributes an importance value to each feature for a particular prediction, representing its contribution to pushing the model's output from the base value to the final prediction. It provides a unified framework that connects many other XAI methods.
Link: https://arxiv.org/abs/1705.07874