Learning Interpretable Concept Groups in CNNs
This addresses the challenge of making CNNs more interpretable for researchers and practitioners, though it is incremental as it builds on existing interpretability techniques.
The paper tackles the problem of improving interpretability in CNNs by proposing Concept Group Learning (CGL), a training methodology that partitions filters into concept groups to learn single visual concepts, resulting in increased interpretability scores and more semantically relevant activation regions.
We propose a novel training methodology -- Concept Group Learning (CGL) -- that encourages training of interpretable CNN filters by partitioning filters in each layer into concept groups, each of which is trained to learn a single visual concept. We achieve this through a novel regularization strategy that forces filters in the same group to be active in similar image regions for a given layer. We additionally use a regularizer to encourage a sparse weighting of the concept groups in each layer so that a few concept groups can have greater importance than others. We quantitatively evaluate CGL's model interpretability using standard interpretability evaluation techniques and find that our method increases interpretability scores in most cases. Qualitatively we compare the image regions that are most active under filters learned using CGL versus filters learned without CGL and find that CGL activation regions more strongly concentrate around semantically relevant features.