Design of multi-level gradient the algorithms in bivalent programming problems

Randomization leads to a variational way to design of multilevel adaptive bivalent programming algorithms by using the «conditions of local improvements». The probabilistic approach to solution of optimization problems helps unify the solution algorithms. Experimental results are given.