[Back]

G. Matz:
"Data Science by TV on the Graph";
Talk: Workshop on Mathematical Data Science, Dürnstein (invited); 10-13-2019 - 10-15-2019.

Graph signal processing is a modern paradigm to deal with large data sets. It captures the intrinsic structure of the data via the topology of a graph. By capitalizing on the graph structure, diverse large-scale learning and inference problems can be tackled. Graph signal processing is promising in applications like sensor networks, social networks, infrastructure networks, or biological networks.In this talk I will report some of our recent work in which we build on the notion of graph total variation to formulate a consistent theoretical framework and efficient distributed algorithms for data reconstruction, network structure inference, and clustering.