Research summary
Thanks to new technologies, we can now observe complex systems—both natural and man-made—on multiple time and space scales. But to understand them, we need complex data representations, such as graphs—and data-driven procedures must be able to process sequences of graphs. Current methods fall short, limiting their scope and impact in applications.
Dr. Lorenzo Livi, Canada Research Chair in Complex Data, is developing methodologies for performing change detection on sequences of attributed graphs with variable numbers of vertices and edges. He and his research team are mapping graphs onto numeric vectors to produce bona fide representations of the original graphs, mirroring their statistical properties. Ultimately, this could lead to the development of change-detection algorithms for medical and industrial applications, such as the ability to predict epileptic seizures or detect faults in smart grids.