1 paper
Andong Wang, Wei-Ning Lee, Xiaojuan Qi
To interpret deep networks, one main approach is to associate neurons with human-understandable concepts. However, existing methods often ignore the inherent relationships of diffe…