The task-oriented optimization of bases in recognition problems (Q1572692)

This work is performed within the framework of the algebraic approach to the recognition problem. To construct a basis set of algorithmic operators, an iterative process is proposed such that the problem of joint optimization of the recurrent operator and a correction operation is solved at each step of this process. It is shown that the choice of the basis operator is reduced to the standard problem of finding a subsystem of inequalities of maximal weight. An efficient numerical procedure for constructing a monotone correction operation is described.