Fatemeh Aghaeipoor

Rojina Kashefi, Leili Barekatain, Mohammad Sabokrou et al.

Transformers have had a significant impact on natural language processing and have recently demonstrated their potential in computer vision. They have shown promising results over convolution neural networks in fundamental computer vision tasks. However, the scientific community has not fully grasped the inner workings of vision transformers, nor the basis for their decision-making, which underscores the importance of explainability methods. Understanding how these models arrive at their decisions not only improves their performance but also builds trust in AI systems. This study explores different explainability methods proposed for visual transformers and presents a taxonomy for organizing them according to their motivations, structures, and application scenarios. In addition, it provides a comprehensive review of evaluation criteria that can be used for comparing explanation results, as well as explainability tools and frameworks. Finally, the paper highlights essential but unexplored aspects that can enhance the explainability of visual transformers, and promising research directions are suggested for future investment.

Fatemeh Aghaeipoor, Dorsa Asgarian, Mohammad Sabokrou

Explainability of Deep Neural Networks (DNNs) has been garnering increasing attention in recent years. Of the various explainability approaches, concept-based techniques stand out for their ability to utilize human-meaningful concepts instead of focusing solely on individual pixels. However, there is a scarcity of methods that consistently provide both local and global explanations. Moreover, most of the methods have no offer to explain misclassification cases. Considering these challenges, we present a unified concept-based system for unsupervised learning of both local and global concepts. Our primary objective is to uncover the intrinsic concepts underlying each data category by training surrogate explainer networks to estimate the importance of the concepts. Our experimental results substantiated the efficacy of the discovered concepts through diverse quantitative and qualitative assessments, encompassing faithfulness, completeness, and generality. Furthermore, our approach facilitates the explanation of both accurate and erroneous predictions, rendering it a valuable tool for comprehending the characteristics of the target objects and classes.