60 lines
4.6 KiB
TeX
60 lines
4.6 KiB
TeX
\section{Related work}
|
|
% original work rosenblatt/parzen
|
|
% langsam
|
|
% other approaches Fast Gaussian Transform
|
|
% binned version silverman, scott, härdle
|
|
% -> Fourier transfom
|
|
|
|
|
|
The kernel density estimator is a well known non-parametric density estimator, originally described independently by Rosenblatt \cite{rosenblatt1956remarks} and Parzen \cite{parzen1962estimation}.
|
|
It was subject to extensive research and its theoretical properties are well understood.
|
|
A comprehensive reference is given by Scott \cite{scott2015}.
|
|
Although classified as non-parametric, the KDE depends on two free parameters, the kernel function and its bandwidth.
|
|
The selection of a \qq{good} bandwidth is still an open problem and heavily researched.
|
|
An extensive overview regarding the topic of automatic bandwidth selection is given by \cite{heidenreich2013bandwidth}.
|
|
%However, the automatic selection of the bandwidth is not subject of this work and we refer to the literature \cite{turlach1993bandwidth}.
|
|
|
|
The great flexibility of the KDE makes it suitable for many applications.
|
|
However, this comes at the cost of a slow computation speed.
|
|
%
|
|
The complexity of a naive implementation of the KDE is \landau{MN}, given by $M$ evaluations of $N$ data samples as input size.
|
|
%The complexity of a naive implementation of the KDE is \landau{NM} evaluations of the kernel function, given $N$ data samples and $M$ points of the estimate.
|
|
Various methods have been proposed to reduce the computation time of the KDE.
|
|
|
|
% k-nearest neighbor searching
|
|
An obvious way to speed up the computation is to reduce the number of evaluated kernel functions.
|
|
One possible optimization is based on k-nearest neighbour search, performed on spatial data structures.
|
|
These algorithms reduce the number of evaluated kernels by taking the distance between clusters of data points into account \cite{gray2003nonparametric}.
|
|
|
|
% fast multipole method & Fast Gaus Transform
|
|
Another approach is to reduce the algorithmic complexity of the sum over Gaussian functions, by employing a specialized variant of the fast multipole method.
|
|
The term fast Gauss transform was coined by Greengard \cite{greengard1991fast} who suggested this approach to reduce the complexity to \landau{N+M}.
|
|
% However, the complexity grows exponentially with dimension. \cite{Improved Fast Gauss Transform and Efficient Kernel Density Estimation}
|
|
|
|
% FastKDE, passed on ECF and nuFFT
|
|
Recent methods based on the self-consistent KDE proposed by Bernacchia and Pigolotti \cite{bernacchia2011self} allow to obtain an estimate without any assumptions, \ie{} the kernel and bandwidth are both derived during the estimation.
|
|
They define a Fourier-based filter on the empirical characteristic function of a given dataset.
|
|
The computation time was further reduced by \etal{O'Brien} using a non-uniform fast Fourier transform (FFT) algorithm to efficiently transform the data into Fourier space \cite{oBrien2016fast}.
|
|
|
|
% binning => FFT
|
|
In general, it is desirable to compute the estimate directly from the sample set.
|
|
However, reducing the sample size by distributing the data on an equidistant grid can significantly reduce the computation time, if an approximative KDE is acceptable.
|
|
Silverman \cite{silverman1982algorithm} originally suggested to combine adjacent data points into data bins, which results in a discrete convolution structure of the KDE.
|
|
Allowing to efficiently compute the estimate using a FFT algorithm.
|
|
This approximation scheme was later called binned KDE (BKDE) and was extensively studied \cite{fan1994fast} \cite{wand1994fast} \cite{hall1996accuracy}.
|
|
While the FFT algorithm constitutes an efficient algorithm for large sample sets, it adds an noticeable overhead for smaller ones.
|
|
|
|
The idea to approximate a Gaussian filter using several box filters was first formulated by Wells \cite{wells1986efficient}.
|
|
Kovesi \cite{kovesi2010fast} suggested to use two box filters with different widths to increase accuracy maintaining the same complexity.
|
|
To eliminate the approximation error completely, \etal{Gwosdek} \cite{gwosdek2011theoretical} proposed a new approach called extended box filter.
|
|
|
|
This work highlights the discrete convolution structure of the BKDE and elaborates its connection to digital signal processing, especially the Gaussian filter.
|
|
Accordingly, this results in an equivalence relation between BKDE and Gaussian filter.
|
|
It follows, that the above mentioned box filter approach is also an approximation of the BKDE, resulting in an efficient computation scheme presented within this paper.
|
|
This approach has a lower complexity as comparable FFT-based algorithms and adds only a negligible small error, while improving the performance significantly.
|
|
|
|
|
|
|
|
|
|
|