Artificial Neural Networks for Pattern Recognize Handwritten

Abstract

Backpropagation is a supervised learning method on artificial neural networks. The process of training artificial neural networks that are built will be tested several structures of artificial neural networks, with the hope that the best structure is obtained to produce the output with the smallest error. The evaluation purposes use the same data set to form a neural model with a hidden level and the neuron model with two hidden levels then compare these three models both in terms of classification accuracy and training time. A neural network can be concluded that for each architecture, the neural network architecture used to recognize type 3 is 82,42%, type 2 77,5%, type 1 98,1%.