New method explains AI models by separating feature interactions

Original: Beyond Additive Decompositions: Interpretability Through Separability

Source: arxiv.org ↗