Disentangling Visual Transformers: Patch-level Interpretability for Image Classification
This addresses the need for interpretable models in applications where understanding decisions is crucial, though it is incremental as it modifies existing transformer designs.
The paper tackles the problem of interpretability in visual transformers for image classification by proposing the Hindered Transformer (HiT) architecture, which disentangles patch-level influences and achieves a reasonable trade-off in performance.
Visual transformers have achieved remarkable performance in image classification tasks, but this performance gain has come at the cost of interpretability. One of the main obstacles to the interpretation of transformers is the self-attention mechanism, which mixes visual information across the whole image in a complex way. In this paper, we propose Hindered Transformer (HiT), a novel interpretable by design architecture inspired by visual transformers. Our proposed architecture rethinks the design of transformers to better disentangle patch influences at the classification stage. Ultimately, HiT can be interpreted as a linear combination of patch-level information. We show that the advantages of our approach in terms of explicability come with a reasonable trade-off in performance, making it an attractive alternative for applications where interpretability is paramount.