A Novel Modeling Network Structure and Its Heuristic Learning Performance

In this research, we introduce a novel network, Hybrid Radial Based Function Neural Network (HyrRbfNN), an artificial neural network (ANN) in which a hidden layer of radial based function (RBF) is integrated. The training of weight parameters is accomplished using improved particle swarm optimization (iPSO). In total, the network has four layers. (3 hidden, 1 output) The learning performance of our network has been compared to learning performance of adaptive-network based fuzzy inference systems (ANFIS). Training measurement graphs, output and error surfaces of trained methods are displayed for the both methods. The results has shown that our method provides much better training performance.