A Minimum Output Information Loss Method for Stochastic Model Reduction

On the basis of analyzing the information descriptions of stochastic system output variables, the problem of the linear system model reduction was deal with in the framework of information theory. The basic idea is to minimize the output information loss caused by truncation by eliminating the state variables with the least contribution to system output information on the basis of an equal transformation. Before truncation, the original output matrix should be augmented in according to the so-called pseudo-entropy-preserving principle, and an entropy preserving transformation of the original state is required, so as to get the relationship between the state variables and the output information in the specific coordinate. The derived minimum output information loss (MOIL) method connects with the aggregation method firmly, and it can preserve the stability of the original system. Illustrative example was given