Hopfield type of Artificial Neural Network via Election Algorithm as Heuristic Search method for Random Boolean kSatisfiability

Abstract

This study proposed a hybrid computational model by incorporating Election Algorithm (EA) as a heuristics search technique in a Hopfield type of artificial neural network (HNN). The main objective is to improve the learning phase of Hopfield type artificial neural network (HNN) for optimal Random Boolean kSatisfiability representation for higher-order logic. Many researchers in the area of artificial intelligence (AI), machine learning (ML), and artificial neural networks (ANNs) are motived by the multiple processing units that work together to learn, identify patterns and predict information which provides a powerful mechanism for optimizations/search problems and other decision-making problem. Election algorithm (EA) has been utilized due to the policy of extending the power and rule play by the political parties beyond its borders to seek endorsement from voters. This policy plays an important role in accelerating the learning process of Hopfield type of artificial neural network (HNN) for optimal random Boolean kSatisfiability representation. In this work, a different number of neurons (NN) has been manipulated invalidating the robustness and efficiency of EA in HNN for RANkSAT logical clauses. The proposed model has been compared with other existing model-based the global minima ratio, statistical error accumulations, and time complexity during the learning process. The simulated results generated have been presented in the form of graphs. Based on the result of the study, the proposed HNN-RAN-kSAT-EA demonstrates good agreement with the existing HNN-RANkSAT-ACO but outperformed HNN-RANkSAT-ES in term of statistical measures used in this study.