Reduced estimations of a proper local maximum of a posteriori probability in high dimension identification problems (Q1386886)

In the first part of this paper, necessary information is given about the nonreduced Proper Local Maximum of a posteriori Probability (PLMAP) estimation. Then a theorem on reduced PLMAP-estimation is established. In the third section, the use of the reduced PLMAP-estimation is justified as a linearizing observer for the Kalman filter of complete dimension. The last part of this paper considers a probabilistic criterion about PLMAP existence, and a method for the computation of PLMAP-domain probability boundaries in the reduced case is proposed. This method allows to analyze the a posteriori probability density of the reduced vector of indeterminate parameters.