Unveiling Interpretable Behavior In Two-Way High-Dimensional Clinical Data