On design of adaptive optimization algorithms for pseudo-Boolean functions using the local improvement technique

The paper is concerned with optimization of pseudo-Boolean functions [1]. A randomization- based block-diagram for design of multi-level iterative integer optimization algorithms is proposed with the algorithm using generally only the current data on the values of the goal function. The design relies on the local improvement rules [2, 3] which are the basis for methods of the «feasible direction» type. The design is shown of a correcting algorithm which generally can improve the effectiveness of maximization by additional processing of the data obtained.