Karim, Syed Anayet
(2023)
Multi-objective Hybrid Election Algorithm For Random K Satisfiability In Discrete Hopfield Neural Network.
PhD thesis, Universiti Sains Malaysia.
Abstract
In the current Artificial Neural Network research development, symbolic
logical structure plays a vital role for describing the concept of intelligence. The
existing Discrete Hopfield Neural Network with systematic Satisfiability logical
structures failed to produce non-repetitive final neuron states which tends to local
minima solutions. In this regard, this thesis proposed non-systematic Random k
Satisfiability logic for 3 k , where k generates maximum three types of logical
combinations (k=1,3; k=2,3; k=1,2,3) to report the behaviours of higher-order
multiple logical structures. To analyse the logical combinations of Random k
Satisfiability, this thesis will conduct experimentations with several performance
metrics. The analysis revealed that the k=2,3 combination of Random k Satisfiability
has more consistent interpretation and global solutions compared to the other
combinations. Moreover, the optimal performance of Random k Satisfiability logic can
be achieved by applying an efficient algorithm during the training phase of Discrete
Hopfield Neural Network. One of the major features of an efficient algorithm is to
make a proper balance in the exploration and exploitation strategy. In this regard, this
thesis proposed a hybridized algorithm named Hybrid Election Algorithm that can well
maintain the exploration-exploitation strategy.
Actions (login required)
 |
View Item |
