Тип публикации: статья из журнала
Год издания: 2000
Идентификатор DOI: 10.1007/BF02508628
Аннотация: In the present paper, a training algorithm with nontraditional capabilities and self-adaptation is suggested for a new generation of neural networks with search behavior. The results obtained open up new opportunities for progress in the physics of living systems in the direction of modeling of purposeful search behavior at the insПоказать полностьюect level. They enable systems of adaptive control, considering changes in environmental conditions and in properties of the controllable object in the control process, to be developed. The algorithm allows both traditional supervisory training of a neural network with a priori known answers and finish learning during its functioning in accordance with preassigned criteria when knowledge of required states of neuron network outputs is lacking. © 2000 Kluwer Academic/Plenum Publishers.
Журнал: Russian Physics Journal
Выпуск журнала: Т. 43, № 6
Номера страниц: 483-487
ISSN журнала: 10648887
Место издания: Томск
Издатель: Springer New York Consultants Bureau