Structure Parameter Optimized Kernel Based Online Prediction With a Generalized Optimization Strategy for Nonstationary Time Series

In this paper, sparsification techniques aided online prediction algorithms in a reproducing kernel Hilbert space are studied for nonstationary time series. The as usual consist of the selection structure parameters and weight vector updating. For parameters, dictionary is selected by with selective modeling criteria, symmetric covariance matrix intermittently optimized adaptation evolution strategy (CMA-ES). This intermittent optimization can not only improve structure's flexibility utilizing cross relatedness input variables, but also partly alleviate uncertainty arisen order to sufficiently capture underlying dynamic characteristics prediction-error series, generalized designed sequentially construct updating procedures multiple connection modes. highly flexible effective, it capable enhancing ability adaptively track changing due nonstationarity. Finally, perspective top-level design, we summarize information interaction between network topology regressors inner model parameters. Numerical simulations demonstrate that proposed approach has superior performance