Batman Universitesi Yasam Bilimleri Dergisi
www.yasambilimleridergisi.com
Cilt 1, Sayı 2  Ocak-Haziran 2012  (ISSN: 2147-4877, E-ISSN: 2459-0614)
Necmettin Sezgin, Ramazan Tekin, Abidin Çalışkan

NO Makale Adı
1356118552 Using Hopfield Network in Pattern Recognition

The development of computer technology has increased the use of intelligent
systems in many areas of human life. One of these areas is automatically and
correctly identification of the alphanumeric characters and patterns as well as selection
of the desired objects. Benefiting from the partial characteristics of patterns and
previously stored characters or patterns, Hopfield Network has a characteristic feature
to not only identify but also correct/reconstruct the noisy and destroyed characters
or patterns. In the process the network re-calculates the energy of the character or
pattern after even a small change to the pattern given to the input of the network, and
hence, controls the morphologic transform and forces the noisy pattern to converge
to a previously learned and stored nearest pattern. This simulation continues until
the variation in the energy becomes stable. Intelligent systems, which are built in neural
networks responsible for identifying choosing and processing, have important
roles particularly in robotics. In this study a system consisting of Hopfield network
model was developed for identifying the patterns of alphanumeric characters.