EOCG Model

Published: 9 February 2020| Version 1 | DOI: 10.17632/64drxxf8wz.1
Contributors:
, Sebastian Blanc

Description

This Mendeley Data repository "EOCG Model" contains R code to run Empirical Orthogonal Constraint Generation (EOCG), a model aimed at learning a reduced set of dimensions to solve a Multidimensional Knapsack Problem (MKP). The R code also runs the two models ORIGINAL and HYP. The code has been used in the following research article: T. Setzer and S. M. Blanc, Corrigendum to “Empirical Orthogonal Constraint Generation for Multi-Dimensional 0/1 Knapsack Problems” [European Journal of Operational Research, 282, 1 (2020), 58–70], https://doi.org/10.1016/j.ejor.2019.12.029. Corresponding author: Thomas Setzer, E-mail address: thomas.setzer@ku.de. (DOI of original article: https://doi.org/10.1016/j.ejor.2019.09.016). Computational results presented in the referenced manuscript are available in folder R/Results. Guidance on how to read result files is provided in R/Results/readresults.txt.

Files

Steps to reproduce

To run the code, execute script main.R located in folder R/. The code can be used under the MIT license terms (see file R/license.txt). Further information on configuration and parameterization is included as comments in R/main.R. --- The R code contains routines to load and compute MKP instances. In its current version, it supports - Chu & Beasley (1998) instances, introduced/used in Chu, P. C., Beasley, J. E., 1998. A Genetic Algorithm for the Multidimensional Knapsack Problem. Journal of Heuristics 4, 63–86., and - Glover & Kochenberger (1996) instances, introduced/used in Glover, F., Kochenberger, G. A., 1996. Critical Event Tabu Search for Multidimensional Knapsack Problems. In: Osman, I. H., Kelly, J. P. (Eds.), Meta-Heuristics. Springer USA, pp. 407–427. Both sets of problem instances are computed in the above referenced manuscript. To load MKP instances, the functions get_problems() and load_problem() in R/ProblemLoader.R are used. The function get_problems() expects an argument with one of the following values: - "cb5": 05-dimensional Chu & Beasley MKP instances - "cb10": 10-dimensional Chu & Beasley MKP instances - "cb30": 30-dimensional Chu & Beasley MKP instances - "gk": Glover & Kochenberger MKP instances - "all": All Chu & Beasley and Glover & Kochenberger instances. For other sets of problems, the function needs to be extended. The function expects the problem data formatted as in the data archive Drake, J. 2015. Benchmark instances for the Multidimensional Knapsack Problem. 10.13140/2.1.3578.9122, available at https://www.researchgate.net/publication/271198281_Benchmark_instances_for_the_Multidimensional_Knapsack_Problem. When using the data archive, the folders in the archive are to be extracted into the R/problem subfolder. Assuming the R working directory to be set to R/ for Chu & Beasley instances, the folder and file structure should look like: R/problem/chubeas/OR5x100-0.25_1.dat R/problem/chubeas/OR5x100-0.25_10.dat ... R/problem/chubeas/OR30x500-0.75_9.dat IMPORTANT: Move all problem files located in archive subfolders of /chubeas to the /chubeas folder except the last file in a subfolder containing sets of problems (OR5x100.dat, OR5x250.dat, OR5x500.dat, OR10x100.dat, OR10x250.dat, OR10x500.dat, OR30x100.dat, OR30x250.dat, OR30x500.dat). For the Glover & Kochenberger instances, the folder and file structure should look like: R/problem/gk/gk01.dat ... R/problem/gk/gk11.dat Computational results presented in the referenced manuscript are available in folder R/Results. Guidance on how to read result files is provided in R/Results/readresults.txt.

Institutions

Katholische Universitat Eichstatt-Ingolstadt

Categories

Combinatorial Optimization, Discrete Optimization, Packing Problem

Licence