Foundations and methods for clustering in time-evolving graphs

Establish a rigorous mathematical definition of clusters or community structures for time-evolving graphs, and develop efficient, robust clustering algorithms together with meaningful metrics for comparing clustering results in settings where communities can merge, split, appear, or disappear.

Background

The paper emphasizes that community detection in time-evolving graphs is substantially more complex than in static graphs because clusters may undergo events such as merging, splitting, appearance, and disappearance.

While supra-Laplacian formulations are promising, their practicality can be limited by size and coupling choices, and there is a need for clear definitions and evaluation metrics tailored to temporal dynamics to guide algorithmic development and assessment.