ISSN: 1813-0410

Author : Abdul Ameer Kadhum, Maha

Training System Artificial Neural Networks for Recognition Numbers

Maha Abdul Ameer Kadhum

journal of kerbala university, Volume 10, Issue 0, Pages 148-159

This research present a multiply connected neural network designed to recognize input images to it which are the numbers (0……9) by using neucontron network with for layers (U1, U2, U3, U4) in addition to input layer (U0) to after training network and produce database related to its which can use to check the performance rate of network , also we will mention in this research to the way that normalize image before inter it to the network and then the research will appear the way to display the output to the network by designing program using Matlab for simulation Artificial Neural Network. These results show that the improved learning algorithm is effective for letting neural networks learn all trained patterns, which cannot be learned by conventional methods.