The Mining and Learning with Graphs at Scale workshop focuses on methods for operating on massive information networks. We begin by highlighting applications of graph-based learning and graph algorithms for a wide range of areas such as detecting fraud and abuse, query clustering and duplication detection, image and multi-modal data analysis, privacy-respecting data mining and recommendation, and experimental design under interference.

The main body of the presentation is divided into three sections:

In our first segment, we cover graph learning and graph building algorithms which we apply to graphs with billions of nodes, and trillions of potential edges. We also discuss similarity ranking over graphs, and the clustering and community detection methods which power numerous industrial applications. This section concludes with a discussion of graph-based semi-supervised learning techniques.

Our second segment covers the application of neural networks to graph structured data through both positional graph embeddings and graph neural networks (GNNs). We present challenges, and recent results from our team on scalable inference algorithms for GNNs, methods for dealing with bias in graph data, and ensemble approaches to representing nodes which allow more modeling flexibility.

Our final segment discusses different techniques for working with massive graphs. We focus on how to take advantage of both single- and multi-machine parallelism to run algorithms on graphs of up to trillions of edges.