Drug Discovery with Graph Neural Networks -- part 3

Explanations Techniques help us understand the model's behaviour. For example, explanation methods are used to visualize certain parts of the image or to see how it reacts to a certain input. It is a well-established field of machine learning that has many different techniques which can be applied to deep learning (e.g. However, there have been only a few attempts to create explanation methods for graph neural networks (GNNs). Most of the "reuse" methods that were developed in deep learning and try to apply them in the graph domain. If you would like to learn more about state-of-the-art research on explainable GNNs, I would highly recommend looking over my previous article.