Ricerc@Sapienza

Many physical phenomena and complex systems can be modeled as networks of interacting entities. Prominent examples include social networks, the web, computer and transportation networks, and the brain. The goal of the EGS project is to leverage techniques from signal processing to the analysis and understanding of such complex objects when viewed as graph signals. A graph signal is a graph with values assigned to its vertex set, in a similar way that an image signal consists of values arranged on a grid. A graph signal may evolve over time through changes in the graph structure, or though changes in the values of the nodes. Graph signals offer a very rich structure for representing the information of complex systems consisting of smaller parts, as well as processes that take place on these systems.

The project will explore the vast research on signal processing and graph theory towards developing new analytical tools for the study of graph signals, in particular graph signals that evolve over time. Building on a sound theoretical basis, the project will revisit classical data mining, knowledge extraction and machine learning tasks through the graph signal lens, with the purpose of developing new models and algorithms. As a practical benchmark for testing our new techniques and methods, we plan to investigate the formulation and evolution of online communites and the automatic detection of events in urban or online settings, as well as the analysis of phenomena on biological networks.