ニューラルネットワークを用いた非線形時系列データのモデリングとその不動点推定の統合化手法

概要

論文の詳細を見る
The saddle type of fixed points are embedded in the chaotic trajectory from complex systems, and the estimation of these points in unknown nonlinear systems is significant. At first, we present a coordinative method to construct a nonlinear model approximating time series from unknown nonlinear system and to estimate its fixed point by using a neural networks and nonlinear optimization techniques.Finally, we apply this presented method to OGY method which was presented as one of the chaos control. We can stabilize chaos to the fixed point by using this method. However, in order to construct feedback rules in OGY method, it is necessary to obtain the system equation and its fixed point explicitly. To overcome this problem, we apply the proposed coordinative method. As a result of this application, we can stabilize chaos to its fixed point by using only the time series from the nonlinear system.We also apply the method to the global optimization method by chaos series. In our method, the global optimum can be estimated by using the chaos dynamics without stabilization by the chaos annealing.
社団法人電気学会の論文
2003-05-01