Unbounded Knapsack Problem: New Results

The recent progress in solving the unbounded knapsack problem (UKP) is tightly related to the phenomenon of dominance which contributes to eliminate the non-prootable object types and reduce drastically the solutions search space. We focus here on the analysis of the computational results obtained by us by (EDUK) for EEcient Dynamic programming for UKP. This new dynamic programming algorithm is based on new and very useful dominance relations between object types. Numerous computational experiments with various data instances are presented to validate our ideas and demonstrate their eeciency. The presented computational experiments give some useful insights concerning the impact of the problem parameters for its hardness. We compare EDUK with the widely used, B&B exact algorithm M T U 2 and observe that it outperforms in most of the cases this specialized B&B package.