FastSHAP: Real-Time Shapley Value Estimation
This addresses the problem of slow Shapley value computation for users needing real-time model explanations, though it is incremental as it builds on existing estimation approaches.
The paper tackles the computational cost of calculating Shapley values for explaining black-box models by introducing FastSHAP, a method that estimates them in a single forward pass using a learned explainer model, achieving high-quality explanations with orders of magnitude speedup.
Shapley values are widely used to explain black-box models, but they are costly to calculate because they require many model evaluations. We introduce FastSHAP, a method for estimating Shapley values in a single forward pass using a learned explainer model. FastSHAP amortizes the cost of explaining many inputs via a learning approach inspired by the Shapley value's weighted least squares characterization, and it can be trained using standard stochastic gradient optimization. We compare FastSHAP to existing estimation approaches, revealing that it generates high-quality explanations with orders of magnitude speedup.