Neural metaheuristics for the multidimensional knapsack problem

The multidimensional knapsack problem (MKP) belongs to a very important class of integer optimization problems. In this study, we propose, develop and test metaheuristics based on the neural-networks paradigm for solving the MKP. We show how domain-specific knowledge can be incorporated within the neural-network framework for solving this NP-Hard problem. We provide a mathematical formulation of the algorithm, outline the steps involved and empirically test our approach on standard benchmark problems in the literature. The results are extremely competitive with those by other heuristics and metaheuristics in the literature.