recognition and control
Gennady Alekseevich Ososkov · 2003
A iicw aliproach t,o strricturiiig and of feed- Forward artificial neural netwrirks (ANN) is proposed. I hat lends t,o overcome,in;my shortcomings of multilayer perceptroiJs ad ANN5 wiih radial basis fiiiictions (RRF- nets). A dynamical training algorithm is dcveldped in or- der t,o kcep l,h'e optimd nhber OS neurons in the hidden layrrs and to guarantee thc finiteness of tlic training pro- cedure due to individual traiuing of each rieurou Results of applying of c.he proposed neural neI.wol-k to recognizing froiital inrages of human hcrs look very promising and givi Tis? I,O propose a rionrcxpeusivc security systern. r. I.