20.500.12556/DKUM-10889 RAZPOZNAVANJE ZNAKOV S HOPFIELDOVO MREŽO CHARACTER RECOGNITION WITH HOPFIELD NETWORK V diplomskem delu smo obdelali delovanje Hopfieldove nevronske mreže pri razpoznavanju znakov, ki so predstavljeni kot dvodimenzionalni binarni vzorci. V začetku smo predstavili osnove nevronskih mrež, nato pa smo opisali še zgradbo in delovanje Hopfieldove nevronske mreže. Predstavili smo problematiko lažnih stanj ter opisali metodo razlikovanja med lažnimi atraktorji in naučenimi stanji mreže. Podrobneje smo opisali programsko rešitev Hopfieldovega algoritma, ki smo jo razvili, nato pa smo predstavili rezultate poskusov, ki smo jih opravili s pomočjo implementacije. In this diploma thesis we have studied Hopfield neural network operation and recognition of characters, which are presented as twodimensional binary patterns. First, we have described the basics of neural networks in general, after which we have also considered the structure and operation of the Hopfield neural network. Further, we have presented the problem of false attractors and the method of distinguishing between false attractors and learned states. We have also presented our implementation of the Hopfield neural net algorithm and the results of experiments, conducted with the help of our program. umetna inteligenca Hopfieldova nevronska mreža lažna stanja Hebbovo učenje artificial intelligence Hopfield neural network false attractors Hebbian learning true true false [M. Višić] Slovenski jezik Angleški jezik Diplomsko delo 2009-06-12 11:58:54 2009-06-23 14:34:46 2022-04-11 16:39:08 2022-04-17 03:18:03 2009 0 Maribor 0 0000-00-00 NiDoloceno NiDoloceno NiDoloceno 0000-00-00 0000-00-00 0000-00-00 004.89:004.92(043.2) 13663510 URN:SI:UM:DK:LTUNGRCK UNI_Visic_Marko_2009.pdf UNI_Visic_Marko_2009.pdf 1 B12348B88FDE6253CB404D1581A1923C 44cf96705e311dd6cb64e59f01d55355f0e5ea6c825eb971999d5aa99c2cee89 f4b5e35e-7c02-11eb-bb7a-00155d0001ca 20.500.12556/dkum/b1f95706-37a5-41eb-90ea-3e4f7e49eb4c https://dk.um.si/Dokument.php?lang=slv&id=9267 Fakulteta za elektrotehniko, računalništvo in informatiko 0 0 0