TRAINING OF CONVOLUTION PRINCIPLE IN VECTOR OPTIMIZATION PROBLEMS.

Adaptive solution algorithms for vector optimization problems are considered. The structure of the algorithms is chosen on the basis of a priori information regarding a particular convolution principle for the quality figures of the solution. The chosen structure is refined in the adaptation process with allowance for current information coming from the decision maker. This refinement is in accordance with local-improvement algorithms, these being probabilistic analogs of methods of feasible directions.