Hilbert basis (linear programming): Difference between revisions
en>Maxal |
en>Mark viking moved wl to earlier instance, also disambiguated and removed tag |
||
| Line 1: | Line 1: | ||
A '''whitening transformation''' is a [[decorrelation]] transformation that transforms a set of random variables having a known [[covariance matrix]] <math>M</math> into a set of new random variables whose covariance is the [[identity matrix]] (meaning that they are [[uncorrelated]] and all have [[variance]] 1). | |||
The transformation is called "whitening" because it changes the input vector into a [[white noise|white noise vector]]. It differs from a general decorrelation transformation in that the latter only makes the covariances equal to zero, so that the correlation matrix may be any diagonal matrix. | |||
The inverse [[coloring transformation]] transforms a vector <math>Y</math> of uncorrelated variables (a white random vector) into a vector <math>X</math> with a specified covariance matrix. | |||
== Definition == | |||
Suppose <math>X</math> is a [[random vector|random (column) vector]] with covariance matrix <math>M</math> and mean 0. The matrix <math>M</math> can be written as the expected value of the [[outer product]] of <math>X</math> with itself, namely: | |||
:<math> M = \operatorname{E}[X X^\top]</math> | |||
Since <math>M</math> is [[symmetric matrix|symmetric]] and [[Positive-semidefinite matrix|positive semidefinite]], it has a [[square root of a matrix|square root]] <math>M^{1/2}</math>, such that <math> M^{1/2}( M^{1/2})^\top = M</math>. If <math>M</math> is positive definite, <math>M^{1/2}</math> is invertible. Then the vector <math>Y = M^{-1/2}X</math> has covariance matrix: | |||
<math> \operatorname{Cov}(Y) = \operatorname{E}[Y Y^\top] = M^{-1/2} \operatorname{E}[X X^\top] (M^{-1/2})^\top </math> | |||
<math> \operatorname{Cov}(Y)= M^{-1/2} M (M^{-1/2})^\top = (M^{-1/2} M^{1/2}) ((M^{1/2})^\top (M^{-1/2})^\top ) = I</math> | |||
and is therefore a white random vector. | |||
Since the square root of a matrix is not unique, the whitening transformation is not unique either. | |||
If <math>M</math> is not positive definite, then <math>M^{1/2}</math> is not invertible, and it is impossible to map <math>X</math> to a white vector of the same length. In that case the vector <math>X</math> can still be mapped to a smaller white vector <math>Y</math> with <math>m</math> elements, where <math>m</math> is the number of non-zero [[eigenvalue]]s of <math>M</math>. | |||
==See also== | |||
* [[Decorrelation]] | |||
* [[Randomness extractor]] | |||
* [[Hardware random number generator]] | |||
* [[Principal component analysis]] | |||
==References== | |||
{{reflist}} | |||
== External links == | |||
* http://courses.media.mit.edu/2010fall/mas622j/whiten.pdf | |||
[[Category:Classification algorithms]] | |||
Latest revision as of 22:01, 13 September 2013
A whitening transformation is a decorrelation transformation that transforms a set of random variables having a known covariance matrix into a set of new random variables whose covariance is the identity matrix (meaning that they are uncorrelated and all have variance 1).
The transformation is called "whitening" because it changes the input vector into a white noise vector. It differs from a general decorrelation transformation in that the latter only makes the covariances equal to zero, so that the correlation matrix may be any diagonal matrix.
The inverse coloring transformation transforms a vector of uncorrelated variables (a white random vector) into a vector with a specified covariance matrix.
Definition
Suppose is a random (column) vector with covariance matrix and mean 0. The matrix can be written as the expected value of the outer product of with itself, namely:
Since is symmetric and positive semidefinite, it has a square root , such that . If is positive definite, is invertible. Then the vector has covariance matrix:
and is therefore a white random vector.
Since the square root of a matrix is not unique, the whitening transformation is not unique either.
If is not positive definite, then is not invertible, and it is impossible to map to a white vector of the same length. In that case the vector can still be mapped to a smaller white vector with elements, where is the number of non-zero eigenvalues of .
See also
References
43 year old Petroleum Engineer Harry from Deep River, usually spends time with hobbies and interests like renting movies, property developers in singapore new condominium and vehicle racing. Constantly enjoys going to destinations like Camino Real de Tierra Adentro.