The analysis of Tabu Machine parameters applied to discrete optimization problems

In this paper, we set out results of the comparative investigation of the neural network approach for the discrete optimization problems in a case of Tabu search usage. The discussed neural networks are known as Tabu Machine, which consists of a set of binary nodes connected between each other by bi-directional links. They change states by a predefined neural network algorithm in order to find a global minimum of the network energy. The problem of the optimal logical structure synthesis for distributed databases is chosen as the test case for the investigation. In the course of the research, we obtained the guidelines for proper selection of the Tabu Machine parameters space, which may be used either to improve quality of the solution, or also to increase efficiency of its finding in comparison with the optimization algorithm based on Hopfield networks.