Graphs are a powerful tool to represent data that is produced by a variety of artificial and natural processes. A graph has a compositional nature, being a compound of atomic information pieces, and a relational nature, as the links defining its structure denote relationships between the linked entities. Also, graphs allow us to represent a multitude of associations through link orientation and labels, such as discrete relationship types, chemical properties, and strength of the molecular bonds.
But most importantly, graphs are ubiquitous. In chemistry and material sciences, they represent the molecular structure of a compound, protein inter- action and drug interaction networks, biological and bio-chemical associations. In social sciences, networks are widely used to represent peopleā€™s relationships, whereas they model complex buying behaviors in recommender systems.
Graph machine learning helps to learn and infer from graph data by embedding discrete and continuous features into traditional (flat data) algorithms.

Go to Top