A self-organizing genetic algorithm for finding a global optimum of the goa
l function with unknown relief is proposed. To find a global extremum, a un
ique launch of the algorithm is required. The self-organization principles
are implemented in constructing the algorithm via the competition of variou
s reproduction and mutation schemes, as well as via the adjustment of the n
umber of parents in multiparent reproduction schemes in the course of opera
tion of the algorithm, which provides an improved efficiency of these schem
es. The convergence of the developed algorithm is proved.