A New Tabu Search Approach to the 0–1 Equicut Problem

Given an undirected graph, the 0–1 equicut problem consists of finding a partition of the vertex set into two subsets of equal size, such that the number of edges going from one subset to the other is minimized. A classical heuristics for this problem was presented 25 years ago, whereas simulated annealing, genetic algorithms, tabu search and a greedy randomized procedure have been developed in the last 5 years. In this paper we present a new tabu search algorithm and show, thorough extensive computational experiments, that in most cases it beats the other methods.