One Machine Recognition Problem in Self-training

Machine recognition in self-training mode is formulated as a combinatorial-type extremal problem. It is shown that an exact solution can be obtained by integer linear programming; approximate-solution algorithms are indicated.