J.M. Crane and M. Zhang (USA)
Data Simulation, Data Simulator, HONN, and SINCHONN *This research is supported by Christopher Newport University (CNU) 2004 Summer Stipend grant and CNU Applied Research Centre 2004 research funding.
Real world data is rough and intermittent. When applying neural networks to simulate such functions, accuracy is a problem. Higher Order Neural Network (HONN) models have the ability to converge without difficulty. A single model will have the ability to simulate a specific type of data more accurately than the rest. This paper presents the development of a SINC Polynomial Higher Order Neural Network (SINCHONN) model. Development of a new software package used to implement SINCHONN Simulator. Exclusive-OR (XOR) data has been tested to demonstrate that SINCHONN can converge. Performance was based on accuracy and time. Foreign Exchange Rate data has been used to test the new model and found that SINCHONN is 55% better than PHONN, 35% better than THONN, 58% better than SPHONN on specified simulation data. By creating this model, the Neuron-Adaptive Higher Order Neural network Group (NAHONG) can be expanded to have a greater scope and reach a higher level of accuracy.
Important Links:
Go Back
