Search

The problem of solving the task of discrete optimization is not completely solved. Lack of publications, scientific developments, algorithms and software products do not give the ability to transfer the mathematical apparatus of solving discrete optimization problems to the class P of computational complexity. All analytical and combinatorial algorithms for solving problems of linear and non-linear optimization are NP completeness. There are developments to improve the efficiency of robotic algorithms, they are filled with requests and actual ones. In this paper, it is proposed to use deterministic and probabilistic methods for forming a priority queue of decision vector components in order to assign positive values to them. After the formation of the variant of the solution, it is possible to vectorize as if the approached solution is taken away from the record value of the goal function, which in exact algorithms is like a solution, which results in a polyp. In this paper has a method for forming a priority line for concretizing the components of the solution vector. The basis of deterministic methods is the idea of a greedy algorithm. The location in the queue is determined by the value of the corresponding component of the cost vector. The appearance of the value of non-visibility in the system of restrictions increases the priority of the component. Behind such a way, another method of determination is modified. Probability assessment probability of priorities is based on the ideas of algorithms in an ant colony and simulated annealing. The scope of probability indicates the significance of the component - a contender for a positive value. A numerical example of a small variability of the task about a knapsack has been introduced, which demonstrates the imitation of a nearby solution. [ABSTRACT FROM AUTHOR]