Enhanced Hopfield Neural Networks With Artificial Immune System Algorithm For Satisfiability Logic Programming
Loading...
Date
2017-08
Authors
Mansor, Mohd. Asyraf
Journal Title
Journal ISSN
Volume Title
Publisher
Universiti Sains Malaysia
Abstract
The emergence of 3-Satisfiability (3-SAT) problem has produced a prolific number
of works devoted to the field of logic and data mining. In this study, a new hybrid
method in doing logic programming by incorporating 3-SAT logical rules as a computational
tool will be presented. Hence, a robust intelligence system that integrates
the Hopfield neural network and metaheuristic paradigm is constructed to extract the
data set hidden knowledge in the form of 3-Satisfiability logical rule. A hybrid network
called HNN-3SATAIS is proposed by assimilating the Hopfield neural network
with the enhanced artificial immune system (AIS) algorithm as a training tool in doing
3-Satisfiability logic programming. The performance of the proposed network,
HNN-3SATAIS is compared with a modified genetic algorithm with Hopfield neural
network (HNN-3SATGA) and the exhaustive search with Hopfield neural network
(HNN-3SATES). Theoretically, the proposed HNN-3SATAIS technique is expected to
reduce the complexity of the network due to the systematic searching mechanism. In
addition, HNN-3SATAIS is a robust method since the metaheuristic algorithm will
boost the searching process that will drive to more feasible global solutions. The performances
of the hybrid techniques were tested by using simulated and real data set.
Dev-C++ Version 5.11 for Windows 10 was used as a platform for training, simulating
and validating the performances of the proposed network. Hence, the appraisal
of hybrid computational models was conducted experimentally by using the randomized
3-SAT instances and 15 real data sets archived from the UCI machine learning
repository. 15 real data sets of different sizes are selected from different fields, ranging
from the finances, astronomy to the vigilant diseases data sets.
Description
Keywords
Enhanced hopfield neural networks with , artificial immune system algorithm