Optimal Performance Evaluation Metrics For Satisfiability Logic Representation In Discrete Hopfield Neural Network

The performance measures and the quality assessment of the solutions for Satis fiability logic in Discrete Hopfield Neural Network (DHNN) are significantly dependent on the selection of the optimal performance evaluation metrics. The current performance measures were mostly leveraging the computational time, absolute error, mean squared error, and goodness of fit measures. To assess the learning capability of a neural network model, the optimal performance metrics are adopted in measuring the quality of the solutions and interpretations obtained by the network especially when dealing with the different number of clauses of Satisfiability logic. The core impetus of this study is to investigate the effects of various performance evaluations metrics towards the models performance analysis based on the learning error, similarity analysis, and energy analysis. Overall, the simulation results have revealed the significant impact of various performance evaluation metrics in terms of learning error, energy evaluation, and similarity analysis for k Satisfiability logic in Discrete Hopfield Neural Network, when k = 3 with different complexities. This finding will reveal the ideal performance metrics that comply with Satisfiability logic and neural network model evaluation.