Corelab Seminar

Danai Koutra
Inferring, Summarizing and Mining Multi-source Graph Data

Networks naturally capture a host of real-world interactions, from social interactions and email communication to brain activity. However, graphs are not always directly observed, especially in scientific domains, such as neuroscience, where monitored brain activity is often captured as time series. How can we efficiently infer networks from time series data (e.g., model the functional organization of brain activity as a network) and speed up the network construction process to scale up to millions of nodes and thousands of graphs? Further, what can be learned about the structure of the graph data? How can we automatically summarize a network `conditionally' to its domain, i.e., summarize its most important properties by taking into account the properties of other graphs in that domain (e.g., neuroscience)? In this talk I will present our recent work on scalable algorithms for inferring, summarizing and mining large collections of graph data coming from different sources. I will also discuss applications in various domains, including connectomics and social science.