Article ID: | iaor19961028 |
Country: | Netherlands |
Volume: | 58 |
Issue: | 1 |
Start Page Number: | 43 |
End Page Number: | 54 |
Publication Date: | Mar 1995 |
Journal: | Journal of Computational and Applied Mathematics |
Authors: | Dyer M.E., Walker J., Riha W.O. |
Keywords: | programming: branch and bound |
Dynamic programming and branch-and-bound methodologies are combined to produce a hybrid algorithm for the multiple-choice knapsack problem. Lagrangian duality is used in a computationally efficient manner to compute tight bounds on every active node in the search tree. The use of Lagrangian duality also enables the use of a reduction procedure to reduce the size of the problem for the enumeration phase. Computational experience with up to 200 multiple-choice sets and 20000 zero-one variables is reported. The computational experience indicates that the resulting algorithm is faster than the best published algorithm and is simpler to code.