Major 3 Satisfiability logic in Discrete Hopfield Neural Network integrated with multi-objective Election Algorithm

• Published in: AIMS mathematics Vol. 8; no. 9; pp. 22447 - 22482
• Main Authors: Roslan, Muhammad Aqmar Fiqhi, Zamri, Nur Ezlin, Mansor, Mohd. Asyraf, Kasihmuddin, Mohd Shareduwan Mohd
• Format: Journal Article
• Language: English
• Published: AIMS Press 01.01.2023
• Subjects: discrete hopfield neural network; election algorithm; flexible logical rule; major 3 satisfiability logic; non-systematic logic
• ISSN: 2473-6988; 2473-6988
• DOI: 10.3934/math.20231145

Discrete Hopfield Neural Network is widely used in solving various optimization problems and logic mining. Boolean algebras are used to govern the Discrete Hopfield Neural Network to produce final neuron states that possess a global minimum energy solution. Non-systematic satisfiability logic is popular due to the flexibility that it provides to the logical structure compared to systematic satisfiability. Hence, this study proposed a non-systematic majority logic named Major 3 Satisfiability logic that will be embedded in the Discrete Hopfield Neural Network. The model will be integrated with an evolutionary algorithm which is the multi-objective Election Algorithm in the training phase to increase the optimality of the learning process of the model. Higher content addressable memory is proposed rather than one to extend the measure of this work capability. The model will be compared with different order logical combinations $ k = \mathrm{3, 2} $, $ k = \mathrm{3, 2}, 1 $ and $ k = \mathrm{3, 1} $. The performance of those logical combinations will be measured by Mean Absolute Error, Global Minimum Energy, Total Neuron Variation, Jaccard Similarity Index and Gower and Legendre Similarity Index. The results show that $ k = \mathrm{3, 2} $ has the best overall performance due to its advantage of having the highest chances for the clauses to be satisfied and the absence of the first-order logic. Since it is also a non-systematic logical structure, it gains the highest diversity value during the learning phase.