Speeding Up Local Search Neighborhood Evaluation for A Multi-Dimensional Knapsack Problem

Abstract

Three different techniques are developed for reducing the computational time required to examine the neighborhood in a local search. Each of the techniques is based on storing some form of sorted information on the problem in order to quickly determine the best move to make. Each technique requires varying amounts of memory and pre-processing time. Computational experiments found that all these techniques improved the computational performance of a tabu search heuristic considerably, however the more complex techniques were able to provide orders of magnitude speed increases over a tabu search based on an unordered neighborhood search.