Probability density estimation using data projection

Abstract

Nonparametric estimation of multivariate multimodal probability density is analysed. The projection pursuit density estimator was proposed by J.H. Friedman. Author of this paper proposes the modifications of original Friedman algorithm: employing a kernel density estimator, and a projection index based on Kolmogorov–Smirnov statistic. The efficiency of proposed modifications is analysed using computer simulation technique.