An Efficient MATLAB Algorithm for Graph Partitioning Technical Report

This report describes a graph partitioning algorithm based on spectral factorization that can be implemented very efficiently with just a hand full of MATLAB commands. The algorithm is closely related to the one proposed by Phillips and Kokotović [3] for state-aggregation in Markov chains. The appendix contains a MATLAB script that implements the algorithm. 1 Graph partitioning Consider an undirected graph G = (V, E) with vertex set V and edge set E, together with a positive edgecost function c : E → [0,∞). A k-partition of V is a collection P = {V1, V2, . . . , Vk} of k disjoint subsets of V , whose union equals V . The cost associated with P is defined by C(P) := ∑