Manifold-based Shapley for SAR Recognization Network Explanation
This work addresses interpretability challenges in high-risk, high-cost scenarios like synthetic aperture radar (SAR) recognition, but it appears incremental as it builds on existing Shapley methods with a manifold projection approach.
The paper tackled the problem of Shapley-based explanations being invalid for high-dimensional models due to feature independence assumptions, and introduced a manifold-based Shapley method that projects features into low-dimensional manifolds to produce Fusion-Shap, resulting in improved interpretability for SAR recognition networks.
Explainable artificial intelligence (XAI) holds immense significance in enhancing the deep neural network's transparency and credibility, particularly in some risky and high-cost scenarios, like synthetic aperture radar (SAR). Shapley is a game-based explanation technique with robust mathematical foundations. However, Shapley assumes that model's features are independent, rendering Shapley explanation invalid for high dimensional models. This study introduces a manifold-based Shapley method by projecting high-dimensional features into low-dimensional manifold features and subsequently obtaining Fusion-Shap, which aims at (1) addressing the issue of erroneous explanations encountered by traditional Shap; (2) resolving the challenge of interpretability that traditional Shap faces in complex scenarios.