The magical world of graph convolutions - David Healey

While machine learning has traditionally focused on representing unstructured pieces of data like images and text, there's a lot of extremely important information encoded in the relationships between people and things as well. Whether in social networks, business communication networks, or networks of atoms in molecules, some of the most relevant information to predict the behavior of a person or thing is encoded in who or what they're connected to. One way of representing these relationships is through graph convolutions and message passing neural networks. This talk will be an overview of these methods, how they work and how to use them in prediction tasks.

David Healey is a data scientist contractor for Recursion Pharmaceuticals, where he's does applied machine learning research for drug discovery. He received a PhD in biology from MIT's Center for the Physics of Living Systems in 2015 and has since done ML research for several local tech companies. He's currently traveling around Europe with his wife and two little boys, and when he's back in Utah he likes to fish.

Food will be provided by Google Cloud