A study on development and implementation of shrinkage estimation algorithms for dynamic systems

点击翻译按钮获取中文摘要

In this research we focus on the concept of shrinkage estimation as a novel statistical inference technique in dynamic stochastic systems, which is frequently used in dependability and performance evaluation. We develop a theoretical framework to perform the statistical estimation with more accuracy and cheaper computation cost, than the common approaches such as parametric estimation based on the likelihood information or non-parametric estimation via kernel functions. Especially, we consider the wavelet-based shrinkage estimation, and propose several statistical estimation algorithms for high-dimensional non-stationary stochastic processes. Further, we develop the statistical analysis tool to implement the wavelet-based statistical estimation algorithms, and investigate several application examples in individual research area.