Construction of Efficient Algorithms for the Estimation of Multivariate Probability Densities

摘要

State of the art software has been created for the nonparametric estimation and contour display of densities of dimensions through five. A nonparametric mode finding algorithm has been developed for densities of dimension up to one hundred. An algorithm for data based simulation from high dimensional distributions has been created and employed on behind armour data. In the related area of parameter estimation of stochastic processes, simulation based algorithms have been developed which enable the user to proceed directly from the axioms of the process plus data to the estimation of parameters. (Author)