Optimizing of the microbiological diagnostics system training parameters

Djulgam Saad, V.O. Vostotskyi

The information-extremal algorithm of medical diagnosing system training under images of medical and biological objects is offered. Thus at a system grade level the task of creation of solving rules by optimal in information sense of a partition of signs space on recognition classes is considered. With a view of efficiency increase and reliability of image understanding at examination, that is it is immediate in a diagnosing mode, the parallel-serial algorithm of control tolerances optimization on recognition signs is developed. Application of such algorithm allows to consider the quasioptimal control tolerances received in the course of parallel optimization, as start for serial algorithm. It guarantees a finding in the course of optimization values of entropy criterion of the functional learning efficiency in working (admissible) area determination of its function. The example of implementation of the offered algorithm of diagnosing system training with optimization of control tolerances on recognition signs for image understanding of the microbiological environment of intestines of the person is considered.

Key words: information-extremal algorithm, training, optimization, control tolerances, recognition signs, information criterion, the microbiological environment.