Instruments and Systems: Monitoring, Control, and Diagnostics Annotation << Back Single-layer perceptron training algorithm for diagnostics and control in electronics Medvedev A.I. The article contemplates an image identification problem originating under diagnostic of electronic devices by the example of Arabic figures, Cyrillic and Roman alphabets recognition by neural net training via solution methods for linear inequality systems. It suggests initial values for calculation step and for weight matrix, which considerably decreases the periods on the neural net training stage and increases algorithm convergence rate.