Research Frontiers in Interpretable Machine Learning for Feature Importance Analysis
Application of SHAP, LIME, and other interpretability methods to identify which observational features most strongly drive astronomical source classification decisions.