Core Concepts
The article characterizes the extremal degree enumerators, which are the vertices of the degree enumerator polytope, for complete graphs and complete bipartite graphs. This characterization leads to faster algorithms for degree sequence optimization problems over these graph classes.
Abstract
The article studies the degree sequence optimization problem, which is to find a subgraph of a given graph that maximizes the sum of given functions evaluated at the subgraph degrees. The authors introduce the concept of degree enumerators, which are vectors representing the number of vertices of each degree in a graph, and the corresponding degree enumerator polytope.
The key insights are:
For complete graphs Kn, the vertices of the degree enumerator polytope En are precisely the degree enumerators of certain regular and almost-regular graphs Gn(r) and Gn(r,s). This characterization leads to a much faster O(n^2) algorithm for degree sequence optimization over Kn.
For complete bipartite graphs Km,n, the vertices of the degree bi-enumerator polytope Bm,n have a more complicated structure. The article completely characterizes the vertices for the case m=2 and any n. This again leads to a faster O(n^2) algorithm for degree sequence bi-optimization over K2,n.
The article also shows that the total number of degree enumerators grows exponentially with n, while the number of extremal enumerators, i.e., the vertices of the polytopes, only grows quadratically.