2 Satisfiability Logic Programming in Radial Basis Function Neural Networks.

The training of a radial basis function neural network (RBFNN) involves finding the optimal number of hidden neurons in the hidden layer and finding the RBFNN parameters such as center, width, and the output weight. 2 Satisfiability logic programming will be embedded in RBFNN during the training phase. Two training techniques, notraining, and half-training are proposed in this paper. The experiment of both techniques has been examined by using Microsoft Visual Studio 2008 C# Express software. The detailed comparison of the performance of two different techniques in performing 2SAT is discussed in term of root mean square error (RMSE), the number of the hidden neurons and CPU time. The results obtained from the computer simulation have shown that RBFNN-2SAT in halftraining technique outperforms than RBFNN-2SAT in no-training technique due to the terms of RMSE, the number of hidden neurons and CPU time are typically much less than the number of data points, and the centers are not restricted to be data points. [ABSTRACT FROM AUTHOR]