arXiv++ Combinatorics

The Maximum Cut (Max-Cut) problem could be naturally expressed either in a Quadratic Unconstrained Binary Optimization (QUBO) formulation, or as an Ising model. It has long been known that the Maximum Independent Set (MIS) problem could also be related to a specific Ising model. Therefore, it would be natural to attack MIS with various Max-Cut/Ising solvers. It turns out that this strategy greatly improves the approximation for the independence number of random Erdős-Rényi graphs. It also exhibits perfect performance on a benchmark arising from coding theory. These results pave the way for further development of approximate quantum algorithms on MIS, and specifically on the corresponding coding problems.

A cut of a graph can be represented in many different ways. Here we propose to represent a cut through a ``relation tree'', which is a spanning tree with signed edges. We show that this picture helps to classify the main greedy heuristics for the maximum cut problem, in analogy with the minimum spanning tree problem. Namely, all versions of the Sahni-Gonzalez~(SG) algorithms could be classified as the Prim class, while various Edge-Contraction~(EC) algorithms are of the Kruskal class. We further elucidate the relation of this framework to the stabilizer formalism in quantum computing, and point out that the recently proposed \textit{ADAPT-Clifford} algorithm is a reformulation of a refined version of the SG algorithm, SG3. Numerical performance of the typical algorithms from the two classes are studied with various kinds of graphs. It turns out that, the Prim-class algorithms perform better for general dense graphs, and the Kruskal-class algorithms performs better when the graphs are sparse enough.

We apply the stabilizer formalism to the Maximum Cut problem, and obtain a new greedy construction heuristic. It turns out to be an elegant synthesis of the edge-contraction and differencing edge-contraction approaches. Utilizing the relation between the Maximum Cut problem and the Ising model, the approximation ratio of the heuristic is easily found to be at least $1/2$. Moreover, numerical results show that the heuristic has very nice performance for graphs with about 100 vertices.