Shapley Computations Using Surrogate Model-Based Trees
This addresses a bottleneck in interpretable machine learning for practitioners needing efficient Shapley computations, though it appears incremental as an approximation method.
The paper tackles the computational expense of computing Shapley and SHAP values for model interpretation by proposing a surrogate model-based tree algorithm, which simulation studies show improves accuracy and allows a trade-off between running time and accuracy.
Shapley-related techniques have gained attention as both global and local interpretation tools because of their desirable properties. However, their computation using conditional expectations is computationally expensive. Approximation methods suggested in the literature have limitations. This paper proposes the use of a surrogate model-based tree to compute Shapley and SHAP values based on conditional expectation. Simulation studies show that the proposed algorithm provides improvements in accuracy, unifies global Shapley and SHAP interpretation, and the thresholding method provides a way to trade-off running time and accuracy.