Coarse-to-Fine Concept Bottleneck Models
About
Deep learning algorithms have recently gained significant attention due to their impressive performance. However, their high complexity and un-interpretable mode of operation hinders their confident deployment in real-world safety-critical tasks. This work targets ante hoc interpretability, and specifically Concept Bottleneck Models (CBMs). Our goal is to design a framework that admits a highly interpretable decision making process with respect to human understandable concepts, on two levels of granularity. To this end, we propose a novel two-level concept discovery formulation leveraging: (i) recent advances in vision-language models, and (ii) an innovative formulation for coarse-to-fine concept selection via data-driven and sparsity-inducing Bayesian arguments. Within this framework, concept information does not solely rely on the similarity between the whole image and general unstructured concepts; instead, we introduce the notion of concept hierarchy to uncover and exploit more granular concept information residing in patch-specific regions of the image scene. As we experimentally show, the proposed construction not only outperforms recent CBM approaches, but also yields a principled framework towards interpetability.
Related benchmarks
| Task | Dataset | Result | Rank | |
|---|---|---|---|---|
| Image Classification | ImageNet (test) | Top-1 Accuracy78.5 | 299 | |
| Image Classification | ImageNet-100 (test) | Clean Accuracy87.3 | 189 | |
| Image Classification | CIFAR10 | Accuracy96.35 | 137 | |
| Image Classification | Places365 | Accuracy48.55 | 79 | |
| Image Classification | CIFAR100 | Accuracy82.33 | 50 | |
| Image Classification | CUB | Accuracy79.56 | 24 | |
| Image Classification | CIFAR100 (test) | Accuracy60.02 | 14 | |
| Image Classification | AWA2 (test) | Accuracy89.19 | 12 | |
| Image Classification | ImageNet | Accuracy78.45 | 10 | |
| Image Classification | CUB | Accuracy79.5 | 10 |