Determining the connectedness of an undirected graph

Abstract

Determining the connectedness of an undirected graph is a frequent issue in practical graph mining and regarded as a key subproblem of the graph partitioning problem. Apart from graph partitioning, graph connectedness also plays an imperative role in tracking the spread of disease, VLSI design, social network analysis, and theoretical studies in graph theory such as 'Cayley graph'. This work reviews several important methods for determining the connectedness of an undirected graph, such as breadth-first search, depth-first search, and the eigenvalues of a graph Laplacian matrix. In addition, we propose several new methods, such as power sum and logical sum of adjacency matrix. We compare all the relevant methods empirically on random graphs with up to 10 000 vertices, and show that the breadth-first search and logical sum methods deliver good performances on large graphs with more than 100 vertices and the logical sum method is the fastest.