Empirically Adjusted Greedy Algoriths (EAGH): A new approach to solving combinatorial optimisation problems

Abstract

A greedy heuristics to solve a given combinatorial optimisation problem can be seen as an element of an infinite set of heuristics, H, which is defined by a function that depends on several parameters. We propose a procedure for determining the best element of H for a set of instances of the combinatorial optimisation problem. The procedure consists essentially in applying a direct non-linear optimization algorithm to a function of the parameters that characterise H.