Artificial Immune System Algorithm for Training Symbolic Radial Basis Function Neural Network Based 2 Satisfiability Logic Programming

Abstract

The process of radial basis function neural network based 2 Satisfiability logic programming (RBFNN-2SAT) depends mainly upon an adequate obtain the linear optimal output weights with the lowest iteration error. In this paper, the capability and effectiveness of the Artificial Immune System algorithm beside RBFNN-2SAT approach are investigated to improve the linear output by find the best output weights. In this paper, AIS algorithm is presented for enhancing the weights during training RBFNN-2SAT. The performance analysis of the presented technique RBFNN-2SATAIS is compared to three techniques, namely the no-training method, which is incorporated with radial basis function neural network 2SAT (RBFNN-2SATNT), the half-training method, which is incorporated into radial basis function neural network 2SAT (RBFNN-2SATHT), in addition to a genetic algorithm incorporated into radial basis function neural network 2SAT (RBFNN-2SATGA). The simulated results showed the paradigm performance vis-à-vis mean absolute error (MAE) and Root Mean Square Error (RMSE), as well as Schwarz Bayesian Criterion (SBC), along with the CPU Time. Accordingly, the introduced approach, i.e., RBFNN-2SATAIS outperformed the corresponding conventional approaches regarding robustness, accuracy, as well as sensitivity throughout simulation. The simulation established that the artificial immune system algorithm has effectively complied in tandem with radial basis function neural network 2SAT.