Graphs are pretty much all around us — molecules, transportation networks, social networks, even the connectome in the human brain.
In many ways, graphs are the main modality of data we receive from nature. While machine learning has made its most significant strides on data that can be neatly organised into a grid — such as images or text — very rarely will naturally occurring information be directly mappable to such a structure.
Accordingly, graph representation learning is likely a key component towards many scientific and industrial applications of Artificial Intelligence. Advances in graph neural networks (GNNs) and related techniques have led to new state-of-the-art results in numerous domains: chemical synthesis, vehicle routing, 3D-vision, recommender systems, question answering, continuous control, self-driving and social network analysis.
Accordingly, GNNs regularly top the charts on fastest-growing trends and workshops at virtually all top machine learning conferences.
I will give a bird’s eye overview of GNNs, and outline ways in which industrial and scientific players are already having impact with them. Diving deeper, I will expose how we successfully deployed GNNs in production at Google Maps, serving daily travel-time queries worldwide. Lastly, throughout the talk I will call out explicit pointers to useful libraries and data sets that can empower anyone to make an impact in the area!
The talk will be geared towards a generic computer science audience — no prior knowledge of machine learning is necessary.