Abstract
This paper presents a supervised higher order polynomial neural network which is called Dynamic Ridge Polynomial Neural Network. The network combines the characteristics of higher order and recurrent neural networks. It functionally extends the input space into a higher dimensional space, where linear separability is possible, without suffering from the combinatorial explosion in the number of weights. Furthermore, the presence of the recurrent link expands the network's ability for attractor dynamics and storing information for later use. In order predict the fu-ture trends of the S&P 500 signals, a Real Time Recurrent Learning algorithm was employed in training the network. Extensive simulations for the prediction of five steps ahead were performed on the signals. Experimental results indicate that the Dynamic Ridge Polynomial Neural Network demonstrated advantages in capturing chaotic movement in the signals with an improvement in the profit return, and rapid convergence over the widely known Multilayer Perceptrons.
