|
|
| Line 1: |
Line 1: |
| [[File:von mises fisher.png|thumb|Points sampled from three von Mises-Fisher distributions on the sphere (blue: <math>\kappa=1</math>, green: <math>\kappa=10</math>, red: <math>\kappa=100</math>). The mean directions <math>\mu</math> are shown with arrows.]]
| | I am Oscar and I completely dig that name. Years in the past we moved to North Dakota and I adore each working day residing right here. What I love performing is playing baseball but I haven't made a dime with it. In her professional lifestyle she is a payroll clerk but she's always needed her personal business.<br><br>Feel free to surf to my homepage: [http://www.neweracinema.com/tube/user/KOPR www.neweracinema.com] |
| | |
| In [[directional statistics]], the '''von Mises–Fisher distribution''' is a
| |
| [[probability distribution]] on the <math>(p-1)</math>-dimensional [[sphere]] in <math>\mathbb{R}^{p}</math>. If <math>p=2</math>
| |
| the distribution reduces to the [[von Mises distribution]] on the [[circle]]. | |
| | |
| The probability density function of the von Mises-Fisher distribution for the random ''p''-dimensional unit vector <math>\mathbf{x}\,</math> is given by:
| |
| | |
| :<math>
| |
| | |
| f_{p}(\mathbf{x}; \mu, \kappa)=C_{p}(\kappa)\exp \left( {\kappa \mu^T \mathbf{x} } \right)
| |
| | |
| </math>
| |
| | |
| where <math> \kappa \ge 0, \left \Vert \mu \right \Vert =1 \,</math> and
| |
| the normalization constant <math>C_{p}(\kappa)\, </math> is equal to
| |
| | |
| : <math>
| |
| C_{p}(\kappa)=\frac {\kappa^{p/2-1}} {(2\pi)^{p/2}I_{p/2-1}(\kappa)}. \,
| |
| </math>
| |
| | |
| where <math> I_{v}</math> denotes the modified [[Bessel function]] of the first kind and order <math>v</math>. If <math>p=3</math>, the normalization constant reduces to
| |
| : <math>
| |
| C_{3}(\kappa)=\frac {\kappa} {4\pi\sinh \kappa}=\frac {\kappa} {2\pi(e^{\kappa}-e^{-\kappa})}. \,
| |
| </math>
| |
| | |
| Note that the equations above apply for polar coordinates only.
| |
| | |
| The parameters <math>\mu\,</math> and <math>\kappa\,</math> are called the ''mean direction'' and ''[[concentration parameter]]'', respectively. The greater the value of <math>\kappa\,</math>, the higher the concentration of the distribution around the mean direction <math>\mu\,</math>. The distribution is [[unimodal]] for <math>\kappa>0\,</math>, and is uniform on the sphere for <math>\kappa=0\,</math>.
| |
| | |
| The von Mises-Fisher distribution for <math>p=3</math>, also called the Fisher distribution, was first used to model the interaction of dipoles in an electric field (Mardia, 2000). Other applications are found in [[geology]], [[bioinformatics]], and [[text mining]].
| |
| | |
| ==Estimation of parameters==
| |
| A series of <math>N</math> [[independence (probability theory)|independent]] measurements <math>x_i</math> are drawn from a von Mises-Fisher distribution. Define
| |
| | |
| : <math>
| |
| A_{p}(\kappa)=\frac {I_{p/2}(\kappa)} {I_{p/2-1}(\kappa)} . \,
| |
| </math>
| |
| | |
| Then (Sra, 2011) the [[maximum likelihood]] estimates of <math>\mu\,</math> and <math>\kappa\,</math> are given by
| |
| | |
| :<math>
| |
| \mu = \frac{\sum_i^N x_i}{||\sum_i^N x_i||} ,
| |
| </math>
| |
| :<math>
| |
| \kappa = A_p^{-1}(\bar{R}) .
| |
| </math>
| |
| Thus <math>\kappa\,</math> is the solution to
| |
| :<math>
| |
| A_p(\kappa) = \frac{||\sum_i^N x_i||}{N} = \bar{R} .
| |
| </math>
| |
| A simple approximation to <math>\kappa</math> is
| |
| :<math>
| |
| \hat{\kappa} = \frac{\bar{R}(p-\bar{R}^2)}{1-\bar{R}^2} ,
| |
| </math>
| |
| but a more accurate measure can be obtained by iterating the Newton method a few times
| |
| :<math>
| |
| \hat{\kappa}_1 = \hat{\kappa} - \frac{A_p(\hat{\kappa})-\bar{R}}{1-A_p(\hat{\kappa})^2-\frac{p-1}{\hat{\kappa}}A_p(\hat{\kappa})} ,
| |
| </math>
| |
| :<math>
| |
| \hat{\kappa}_2 = \hat{\kappa}_1 - \frac{A_p(\hat{\kappa}_1)-\bar{R}}{1-A_p(\hat{\kappa}_1)^2-\frac{p-1}{\hat{\kappa}_1}A_p(\hat{\kappa}_1)} .
| |
| </math>
| |
| | |
| ==See also==
| |
| * [[Kent distribution]], a related distribution on the two-dimensional unit sphere
| |
| * [[von Mises distribution]], von Mises–Fisher distribution where p=2, the one-dimensional unit circle
| |
| * [[Bivariate von Mises distribution]]
| |
| * [[Directional statistics]]
| |
| | |
| ==References==
| |
| * Dhillon, I., Sra, S. (2003) "Modeling Data using Directional Distributions". Tech. rep., University of Texas, Austin.
| |
| * Fisher, RA, "Dispersion on a sphere'". (1953) ''Proc. Roy. Soc. London Ser. A.'', 217: 295-305
| |
| * {{cite book |title=Directional Statistics |last=Mardia |first=Kanti |authorlink=Kantilal Mardia |coauthors=Jupp, P. E.|year=1999 |publisher=Wiley |isbn=978-0-471-95333-3}}
| |
| * {{cite doi| 10.1007/s00180-011-0232-x}} [http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.186.1887&rep=rep1&type=pdf Preprint]
| |
| | |
| {{ProbDistributions|directional}}
| |
| | |
| {{DEFAULTSORT:Von Mises-Fisher distribution}}
| |
| [[Category:Probability distributions]]
| |
| [[Category:Directional statistics]]
| |
| [[Category:Multivariate continuous distributions]]
| |
| [[Category:Exponential family distributions]]
| |
| [[Category:Continuous distributions]]
| |
I am Oscar and I completely dig that name. Years in the past we moved to North Dakota and I adore each working day residing right here. What I love performing is playing baseball but I haven't made a dime with it. In her professional lifestyle she is a payroll clerk but she's always needed her personal business.
Feel free to surf to my homepage: www.neweracinema.com