Fast implementation of the Tukey depth

Abstract
This paper presents two new algorithms for fast calculating the Tukey depth of dimensions . The first algorithm is naive, and can compute \emph{exactly} the Tukey depth of a single point with complexity . Compared to the first, the second algorithm further utilizes the \emph{quasiconcave} of the Tukey depth function, and hence can be implemented with complexity less than . Both of them are of \emph{combinatorial} properties. Experiments show that the proposed algorithms require quite minimum memory and run much faster than the existing procedures.
View on arXivComments on this paper