Machine Learning on Graphs

Machine learning has historically been successful at dealing with structured datasets, such as tabular data. With recent advances, particularly in deep learning, there are now also well-established and powerful methods for working with image, text and speech data. However, a lot of real-world data does not easily fit into one of these categories. One important class of such data is network or graph data, which can be used to model concepts such as social networks, transaction flow, computer networks and even molecular interactions. Using graphs we can easily represent and capture the complex interactions and dependencies between objects, but it also opens up a question: How can we apply machine learning to graph structured data?