Couner-Propagation Neural Networks Optimization Based on Rough Set

Abstract

The Couner-Propagation neural networks is weak in convergent speed, will easily sink into local minimum, and its choices of initial weights and thresholds lack sound basis. So, a new optimal algorithm of neural network based on rough set was proposed. The new approach integrates the advantages of the two algorithms; it has good understandability, simple computation and exact accuracy. Then a new algorithm based rough set was put forward and used to optimize the design of neural network weights and threshold. The results of simulation show: the new algorithm can get over the insufficiency of CP, and compared with CP, greatly improve the convergent accuracy and speed, and get a good measurement result.