RECONSTRUCTION OF ORDER PARAMETERS BASED ON AWARD\|PENALTY LEARNING MECHANISM

An analysis of unreasonable factor in construction of order parameters in synergetic approach is presented in this paper. It is proved that the unreasonable factor during the dynamic system can be overcome through reconstruction of order parameters. A way of reconstruction of order parameters based on award\|penalty learning mechanism is proposed, which can figure out a group of linear transformation parameters for order parameters using self\|learning power of synergetic neural networks and award\|penalty learning mechanism. The test on samples from real application shows that the new approach can improve the recognition rate greatly and has bright application prospect. Additionally, in order to guide the selection of parameter δ to obtain the best train performance, the influence of parameter δ of award\|penalty learning mechanism on the performance of training is discussed.