Research of neural network simulators through two training data sets
E Zdravkova, N Nenkov
COMPUTER MODELLING & NEW TECHNOLOGIES 2016 20(1) 12-15
Faculty of Mathematics and Informatics, University of Shumen “Episkop Konstantin Preslavsky” 115, Universitetska St., Shumen 9712, Bulgaria
In the present study our aims is to analyze and test two neural networks simulators - Joone and NeuroPh. This will be accomplished by establishing a neural network of 4 layers constituting a multilayer perceptron with sigmoid links. For the purpose of the study were selected two test data sets, which contain integers. Through sequential training the neural network with each of them and subsequently the test results will be obtained for analysis. The study seeks to show how much these two simulators are similar and how different in their characteristics, what neural networks is suitable to be made by them, what are their advantages and disadvantages, how they can be used interchangeably to give certain desired result.