Matérn Covariance Function
   HOME

TheInfoList



OR:

In
statistics Statistics (from German language, German: ', "description of a State (polity), state, a country") is the discipline that concerns the collection, organization, analysis, interpretation, and presentation of data. In applying statistics to a s ...
, the Matérn covariance, also called the Matérn kernel, is a
covariance function In probability theory and statistics, the covariance function describes how much two random variables change together (their ''covariance'') with varying spatial or temporal separation. For a random field or stochastic process ''Z''(''x'') on a dom ...
used in
spatial statistics Spatial statistics is a field of applied statistics dealing with spatial data. It involves stochastic processes (random fields, point processes), sampling, smoothing and interpolation, regional ( areal unit) and lattice ( gridded) data, poin ...
,
geostatistics Geostatistics is a branch of statistics focusing on spatial or spatiotemporal datasets. Developed originally to predict probability distributions of ore grades for mining operations, it is currently applied in diverse disciplines including pet ...
,
machine learning Machine learning (ML) is a field of study in artificial intelligence concerned with the development and study of Computational statistics, statistical algorithms that can learn from data and generalise to unseen data, and thus perform Task ( ...
,
image analysis Image analysis or imagery analysis is the extraction of meaningful information from images; mainly from digital images by means of digital image processing techniques. Image analysis tasks can be as simple as reading barcode, bar coded tags or a ...
, and other applications of multivariate statistical analysis on
metric space In mathematics, a metric space is a Set (mathematics), set together with a notion of ''distance'' between its Element (mathematics), elements, usually called point (geometry), points. The distance is measured by a function (mathematics), functi ...
s. It is named after the Swedish forestry statistician Bertil Matérn. It specifies the covariance between two measurements as a function of the distance d between the points at which they are taken. Since the covariance only depends on distances between points, it is stationary. If the distance is
Euclidean distance In mathematics, the Euclidean distance between two points in Euclidean space is the length of the line segment between them. It can be calculated from the Cartesian coordinates of the points using the Pythagorean theorem, and therefore is o ...
, the Matérn covariance is also
isotropic In physics and geometry, isotropy () is uniformity in all orientations. Precise definitions depend on the subject area. Exceptions, or inequalities, are frequently indicated by the prefix ' or ', hence '' anisotropy''. ''Anisotropy'' is also ...
.


Definition

The Matérn covariance between measurements taken at two points separated by ''d'' distance units is given by Rasmussen, Carl Edward and Williams, Christopher K. I. (2006
Gaussian Processes for Machine Learning
/ref> : C_\nu(d) = \sigma^2\frac^\nu K_\nu\Bigg(\sqrt\frac\Bigg), where \Gamma is the
gamma function In mathematics, the gamma function (represented by Γ, capital Greek alphabet, Greek letter gamma) is the most common extension of the factorial function to complex numbers. Derived by Daniel Bernoulli, the gamma function \Gamma(z) is defined ...
, K_\nu is the modified
Bessel function Bessel functions, named after Friedrich Bessel who was the first to systematically study them in 1824, are canonical solutions of Bessel's differential equation x^2 \frac + x \frac + \left(x^2 - \alpha^2 \right)y = 0 for an arbitrary complex ...
of the second kind, and ''ρ'' and \nu are positive
parameter A parameter (), generally, is any characteristic that can help in defining or classifying a particular system (meaning an event, project, object, situation, etc.). That is, a parameter is an element of a system that is useful, or critical, when ...
s of the covariance. A
Gaussian process In probability theory and statistics, a Gaussian process is a stochastic process (a collection of random variables indexed by time or space), such that every finite collection of those random variables has a multivariate normal distribution. The di ...
with Matérn covariance is \lceil \nu \rceil-1 times differentiable in the mean-square sense.Santner, T. J., Williams, B. J., & Notz, W. I. (2013). ''The design and analysis of computer experiments.'' Springer Science & Business Media.


Spectral density

The power spectrum of a process with Matérn covariance defined on \mathbb^n is the (''n''-dimensional)
Fourier transform In mathematics, the Fourier transform (FT) is an integral transform that takes a function as input then outputs another function that describes the extent to which various frequencies are present in the original function. The output of the tr ...
of the Matérn covariance function (see Wiener–Khinchin theorem). Explicitly, this is given by : S(f)=\sigma^2\frac\left(\frac + 4\pi^2f^2\right)^.


Simplification for specific values of ''ν''


Simplification for ''ν'' half integer

When \nu = p+1/2,\ p\in \mathbb^+ , the Matérn covariance can be written as a product of an exponential and a polynomial of degree p.Stein, M. L. (1999). ''Interpolation of spatial data: some theory for kriging.'' Springer Series in Statistics.Peter Guttorp & Tilmann Gneiting, 2006. "Studies in the history of probability and statistics XLIX On the Matern correlation family," Biometrika, Biometrika Trust, vol. 93(4), pages 989-995, December. The modified Bessel function of a fractional order is given by Equations 10.1.9 and 10.2.15 as \sqrt K_(z) = \frace^\sum_^n \frac \left( 2z \right) ^ . This allows for the Matérn covariance of
half-integer In mathematics, a half-integer is a number of the form n + \tfrac, where n is an integer. For example, 4\tfrac12,\quad 7/2,\quad -\tfrac,\quad 8.5 are all ''half-integers''. The name "half-integer" is perhaps misleading, as each integer n is its ...
values of \nu to be expressed as C_(d) = \sigma^2\exp\left(-\frac\right)\frac\sum_^p\frac\left(\frac\right)^, which gives: * for \nu = 1/2\ (p=0): C_(d) = \sigma^2\exp\left(-\frac\right), * for \nu = 3/2\ (p=1): C_(d) = \sigma^2\left(1+\frac\right)\exp\left(-\frac\right), * for \nu = 5/2\ (p=2): C_(d) = \sigma^2\left(1+\frac+\frac\right)\exp\left(-\frac\right).


The Gaussian case in the limit of infinite ''ν''

As \nu\rightarrow\infty, the Matérn covariance converges to the squared exponential
covariance function In probability theory and statistics, the covariance function describes how much two random variables change together (their ''covariance'') with varying spatial or temporal separation. For a random field or stochastic process ''Z''(''x'') on a dom ...
: \lim_C_\nu(d) = \sigma^2\exp\left(-\frac\right).


Taylor series at zero and spectral moments

From the basic relation satisfied by the Gamma function \Gamma(z)\Gamma(1-z)=\frac and the basic relation satisfied by the Modified Bessel Function of the second K_(x) = \frac \frac and the definition of the modified Bessel functions of the first I_(x)= \sum_^\infty \frac\left(\frac\right)^, the behavior for d\rightarrow0 can be obtained by the following
Taylor series In mathematics, the Taylor series or Taylor expansion of a function is an infinite sum of terms that are expressed in terms of the function's derivatives at a single point. For most common functions, the function and the sum of its Taylor ser ...
(when \nu is not an integer and bigger than 2): C_\nu(d) = \sigma^2\left(1 + \frac\left(\frac\right)^2 + \frac\left(\frac\right)^4 + \mathcal\left(d^\right)\right),\,\, \nu>2 . When defined, the following spectral moments can be derived from the Taylor series: : \begin \lambda_0 & = C_\nu(0) = \sigma^2, \\ pt\lambda_2 & = -\left.\frac\_ = \frac. \end For the case of \nu\in(0,1)\cup(1,2) , similar
Taylor series In mathematics, the Taylor series or Taylor expansion of a function is an infinite sum of terms that are expressed in terms of the function's derivatives at a single point. For most common functions, the function and the sum of its Taylor ser ...
can be obtained: C_\nu(d) = \sigma^2\left(1 + \frac\left(\frac\right)^2 - \frac\left(\frac\right)^ \left(\frac\right)^ + \mathcal\left(d^\right)\right),\,\, \nu\in (0,1)\cup(1,2) . When \nu is an integer limiting values should be taken, (see ).


See also

*
Radial basis function In mathematics a radial basis function (RBF) is a real-valued function \varphi whose value depends only on the distance between the input and some fixed point, either the origin, so that \varphi(\mathbf) = \hat\varphi(\left\, \mathbf\right\, ), o ...


References

{{DEFAULTSORT:Matern covariance function Geostatistics Spatial analysis Covariance and correlation