Logic Programming In Radial Basis Function Neural Networks
Loading...
Date
2013-11
Authors
Hamadneh, Nawaf
Journal Title
Journal ISSN
Volume Title
Publisher
Universiti Sains Malaysia
Abstract
In this thesis, I established new techniques to represent logic programming in radial
basis function neural networks. Two techniques were developed. The first technique
is to encode the logic programming in radial basis function neural networks. The
second technique is to compute the single step operator of logic programming in radial
basis function neural networks. I used different types of optimization algorithms to
improve the performance of the neural networks. I used three different techniques
for improving the predictive capability of the neural networks. These techniques are:
no-training technique, half training technique and full training technique. In this thesis,
I established a new method for determining the best number of the hidden neurons
in radial basis function neural networks. To do that I used the root mean square
error function and Schwarz bayesian criterion as model selection criteria. I used real
data sets of different sizes in the computational results. The analysis revealed that
performance of particle swarm optimization algorithm and Prey predator algorithm are
better to use in training the networks. In this thesis also, I developed a new technique
to extract the logic programming from radial basis function neural networks. To do
that, I established the radial basis function neural networks which represent the three
conjunctive normal form (3-CNF) logic programming. Following this, I implemented
the results to represent the electronic circuits in the radial basis function neural
networks.
Description
Keywords
Logic Programming , Neural Networks