Explainable Structured Machine Learning in Similarity, Graph and Transformer Models

This course focuses on explainable structured machine learning in similarity, graph, and transformer models. The learning outcomes include understanding explanation techniques within the framework of layer-wise relevance propagation, extending approaches for evaluating and visualizing explanations, and applying these methods to research use cases such as quantifying knowledge evolution and studying gender bias in language models. The course teaches skills in second-order and higher-order attributions and how to improve model faithfulness and generate better explanations. The teaching method involves presentations of research use cases and highlighting the importance of considering model structure in explainable AI. This course is intended for individuals interested in explainable AI, machine learning, and model interpretability.