A genetic programming hyper-heuristic for the multidimensional knapsack problem

Abstract

Abstract-Hyper-heuristics are a class of high-level search techniques which operate on a search space of heuristics rather than directly on a search space of solutions. Early hyperheuristics focussed on selecting and applying a low-level heuristic at each stage of a search. Recent trends in hyper-heuristic research have led to a number of approaches being developed to automatically generate new heuristics from a set of heuristic components. This work investigates the suitability of using genetic programming as a hyper-heuristic methodology to generate constructive heuristics to solve the multidimensional 0-1 knapsack problem. A population of heuristics to rank knapsack items are trained on a subset of test problems and then applied to unseen instances. The results over a set of standard benchmarks show that genetic programming can be used to generate constructive heuristics which yield human-competitive results.

Modified Choice Function Heuristic Selection for the Multidimensional Knapsack Problem

“…This paper investigated the application of a number of Simple Random, Greedy and Choice Function-based hyper-heuristic approaches to a real-world sales summit scheduling problem using two deterministic move acceptance criteria: All Moves and Only Improving. Choice Function heuristic selection has also been used by Bilgin et al [15] for benchmark function optimisation,Özcan et al [16] and Burke et al [17] for examination timetabling and Drake et al [12] for the MKP.…”

“…a selection hyper-heuristic using Late Acceptance Strategy move acceptance criterion [5]). Hyper-heuristics have been applied successfully to a wide range of problems including: production scheduling [6], nurse rostering [7], examination timetabling [7], sports scheduling [8], bin packing [9], dynamic environments [10], vehicle routing [4] and the multidimensional 0-1 knapsack problem [11,12].…”

Sparse, Continuous Policy Representations for Uniform Online Bin Packing via Regression of Interpolants

“…These methods seek to generate new heuristics, operating over a search space of heuristics rather than directly over a space of solutions (e.g. [7,8]). A number of generative hyper-heuristic approaches exist in the online bin-packing literature, with previous work focussing on generating packing policies using different representations.…”

Automatically Designing More General Mutation Operators of Evolutionary Programming for Groups of Function Classes Using a Hyper-Heuristic

“…For an offline hyper-heuristic collects knowledge, from a training a set of instances to solve unknown instances of the same problem. Recently GP has been used with hyperheuristics for the bin packing problem [6], the multidimensional knapsack problem [8], to evolve highly competitive general algorithms for envelope reduction in sparse matrices [12], to handle multiple conflicting objectives in dynamic job shop scheduling [18], to automatically design a mutation operator for Evolutionary Programming [10], to compare rule representations [11], to evolve due-date assignment models in job shop environments [20], to automatic design schedule policies for dynamic multi-objective job shop scheduling [19], to evolve ensembles of dispatching rules for the job shop scheduling problem [22], for feature selection and questionanswer ranking in IBM Watson [2], to automated design production scheduling heuristics [3].…”