Less is More: Fewer Interpretable Region via Submodular Subset Selection

Open in new window