In
linear algebra
Linear algebra is the branch of mathematics concerning linear equations such as:
:a_1x_1+\cdots +a_nx_n=b,
linear maps such as:
:(x_1, \ldots, x_n) \mapsto a_1x_1+\cdots +a_nx_n,
and their representations in vector spaces and through matrices.
...
, the singular value decomposition (SVD) is a
factorization
In mathematics, factorization (or factorisation, see American and British English spelling differences#-ise, -ize (-isation, -ization), English spelling differences) or factoring consists of writing a number or another mathematical object as a p ...
of a
real
Real may refer to:
Currencies
* Brazilian real (R$)
* Central American Republic real
* Mexican real
* Portuguese real
* Spanish real
* Spanish colonial real
Music Albums
* ''Real'' (L'Arc-en-Ciel album) (2000)
* ''Real'' (Bright album) (2010) ...
or
complex
Complex commonly refers to:
* Complexity, the behaviour of a system whose components interact in multiple ways so possible interactions are difficult to describe
** Complex system, a system composed of many components which may interact with each ...
matrix
Matrix most commonly refers to:
* ''The Matrix'' (franchise), an American media franchise
** ''The Matrix'', a 1999 science-fiction action film
** "The Matrix", a fictional setting, a virtual reality environment, within ''The Matrix'' (franchis ...
. It generalizes the
eigendecomposition
In linear algebra, eigendecomposition is the factorization of a matrix into a canonical form, whereby the matrix is represented in terms of its eigenvalues and eigenvectors. Only diagonalizable matrices can be factorized in this way. When the matr ...
of a square
normal matrix In mathematics, a complex square matrix is normal if it commutes with its conjugate transpose :
The concept of normal matrices can be extended to normal operators on infinite dimensional normed spaces and to normal elements in C*-algebras. As ...
with an orthonormal eigenbasis to any
matrix. It is related to the
polar decomposition
In mathematics, the polar decomposition of a square real or complex matrix A is a factorization of the form A = U P, where U is an orthogonal matrix and P is a positive semi-definite symmetric matrix (U is a unitary matrix and P is a positive se ...
.
Specifically, the singular value decomposition of an
complex matrix is a factorization of the form
where is an
complex
unitary matrix
In linear algebra, a complex square matrix is unitary if its conjugate transpose is also its inverse, that is, if
U^* U = UU^* = UU^ = I,
where is the identity matrix.
In physics, especially in quantum mechanics, the conjugate transpose is ...
,
is an
rectangular diagonal matrix
In linear algebra, a diagonal matrix is a matrix in which the entries outside the main diagonal are all zero; the term usually refers to square matrices. Elements of the main diagonal can either be zero or nonzero. An example of a 2×2 diagonal m ...
with non-negative real numbers on the diagonal, is an
complex unitary matrix, and
is the
conjugate transpose
In mathematics, the conjugate transpose, also known as the Hermitian transpose, of an m \times n complex matrix \boldsymbol is an n \times m matrix obtained by transposing \boldsymbol and applying complex conjugate on each entry (the complex con ...
of . Such decomposition always exists for any complex matrix. If is real, then and can be guaranteed to be real
orthogonal
In mathematics, orthogonality is the generalization of the geometric notion of ''perpendicularity''.
By extension, orthogonality is also used to refer to the separation of specific features of a system. The term also has specialized meanings in ...
matrices; in such contexts, the SVD is often denoted
The diagonal entries
of
are uniquely determined by and are known as the
singular value In mathematics, in particular functional analysis, the singular values, or ''s''-numbers of a compact operator T: X \rightarrow Y acting between Hilbert spaces X and Y, are the square roots of the (necessarily non-negative) eigenvalues of the self ...
s of . The number of non-zero singular values is equal to the
rank
Rank is the relative position, value, worth, complexity, power, importance, authority, level, etc. of a person or object within a ranking, such as:
Level or position in a hierarchical organization
* Academic rank
* Diplomatic rank
* Hierarchy
* H ...
of . The columns of and the columns of are called left-singular vectors and right-singular vectors of , respectively. They form two sets of
orthonormal bases
In mathematics, particularly linear algebra, an orthonormal basis for an inner product space ''V'' with finite dimension is a basis for V whose vectors are orthonormal, that is, they are all unit vectors and orthogonal to each other. For example, ...
and , and if they are sorted so that the singular values
with value zero are all in the highest-numbered columns (or rows), the singular value decomposition can be written as
where
is the rank of .
The SVD is not unique. It is always possible to choose the decomposition so that the singular values
are in descending order. In this case,
(but not and ) is uniquely determined by .
The term sometimes refers to the compact SVD, a similar decomposition
in which
is square diagonal of size
, where
is the rank of , and has only the non-zero singular values. In this variant, is an
semi-unitary matrix and
is an
semi-unitary matrix, such that
Mathematical applications of the SVD include computing the
pseudoinverse
In mathematics, and in particular, algebra, a generalized inverse (or, g-inverse) of an element ''x'' is an element ''y'' that has some properties of an inverse element but not necessarily all of them. The purpose of constructing a generalized inv ...
, matrix approximation, and determining the rank,
range
Range may refer to:
Geography
* Range (geographic), a chain of hills or mountains; a somewhat linear, complex mountainous or hilly area (cordillera, sierra)
** Mountain range, a group of mountains bordered by lowlands
* Range, a term used to i ...
, and
null space
In mathematics, the kernel of a linear map, also known as the null space or nullspace, is the linear subspace of the Domain of a function, domain of the map which is mapped to the zero vector. That is, given a linear map between two vector space ...
of a matrix. The SVD is also extremely useful in all areas of science,
engineering
Engineering is the use of scientific method, scientific principles to design and build machines, structures, and other items, including bridges, tunnels, roads, vehicles, and buildings. The discipline of engineering encompasses a broad rang ...
, and
statistics
Statistics (from German language, German: ''wikt:Statistik#German, Statistik'', "description of a State (polity), state, a country") is the discipline that concerns the collection, organization, analysis, interpretation, and presentation of ...
, such as
signal processing
Signal processing is an electrical engineering subfield that focuses on analyzing, modifying and synthesizing ''signals'', such as audio signal processing, sound, image processing, images, and scientific measurements. Signal processing techniq ...
,
least squares
The method of least squares is a standard approach in regression analysis to approximate the solution of overdetermined systems (sets of equations in which there are more equations than unknowns) by minimizing the sum of the squares of the res ...
fitting of data, and
process control
An industrial process control in continuous production processes is a discipline that uses industrial control systems to achieve a production level of consistency, economy and safety which could not be achieved purely by human manual control. I ...
.
Intuitive interpretations
Rotation, coordinate scaling, and reflection
In the special case when is an real
square matrix
In mathematics, a square matrix is a matrix with the same number of rows and columns. An ''n''-by-''n'' matrix is known as a square matrix of order Any two square matrices of the same order can be added and multiplied.
Square matrices are often ...
, the matrices and can be chosen to be real matrices too. In that case, "unitary" is the same as "
orthogonal
In mathematics, orthogonality is the generalization of the geometric notion of ''perpendicularity''.
By extension, orthogonality is also used to refer to the separation of specific features of a system. The term also has specialized meanings in ...
". Then, interpreting both unitary matrices as well as the diagonal matrix, summarized here as , as a
linear transformation
In mathematics, and more specifically in linear algebra, a linear map (also called a linear mapping, linear transformation, vector space homomorphism, or in some contexts linear function) is a mapping V \to W between two vector spaces that pre ...
of the space , the matrices and represent
rotations or
reflection Reflection or reflexion may refer to:
Science and technology
* Reflection (physics), a common wave phenomenon
** Specular reflection, reflection from a smooth surface
*** Mirror image, a reflection in a mirror or in water
** Signal reflection, in ...
of the space, while
represents the
scaling
Scaling may refer to:
Science and technology
Mathematics and physics
* Scaling (geometry), a linear transformation that enlarges or diminishes objects
* Scale invariance, a feature of objects or laws that do not change if scales of length, energ ...
of each coordinate by the factor . Thus the SVD decomposition breaks down any linear transformation of into a
composition
Composition or Compositions may refer to:
Arts and literature
*Composition (dance), practice and teaching of choreography
*Composition (language), in literature and rhetoric, producing a work in spoken tradition and written discourse, to include v ...
of three geometrical
transformations: a rotation or reflection (), followed by a coordinate-by-coordinate
scaling
Scaling may refer to:
Science and technology
Mathematics and physics
* Scaling (geometry), a linear transformation that enlarges or diminishes objects
* Scale invariance, a feature of objects or laws that do not change if scales of length, energ ...
(
), followed by another rotation or reflection ().
In particular, if has a positive determinant, then and can be chosen to be both rotations with reflections, or both rotations without reflections. If the determinant is negative, exactly one of them will have a reflection. If the determinant is zero, each can be independently chosen to be of either type.
If the matrix is real but not square, namely with , it can be interpreted as a linear transformation from to . Then and can be chosen to be rotations/reflections of and , respectively; and
, besides scaling the first
coordinates, also extends the vector with zeros, i.e. removes trailing coordinates, so as to turn into .
Singular values as semiaxes of an ellipse or ellipsoid
As shown in the figure, the
singular values In mathematics, in particular functional analysis, the singular values, or ''s''-numbers of a compact operator T: X \rightarrow Y acting between Hilbert spaces X and Y, are the square roots of the (necessarily non-negative) eigenvalues of the self- ...
can be interpreted as the magnitude of the semiaxes of an
ellipse
In mathematics, an ellipse is a plane curve surrounding two focus (geometry), focal points, such that for all points on the curve, the sum of the two distances to the focal points is a constant. It generalizes a circle, which is the special ty ...
in 2D. This concept can be generalized to -dimensional
Euclidean space
Euclidean space is the fundamental space of geometry, intended to represent physical space. Originally, that is, in Euclid's Elements, Euclid's ''Elements'', it was the three-dimensional space of Euclidean geometry, but in modern mathematics ther ...
, with the singular values of any
square matrix
In mathematics, a square matrix is a matrix with the same number of rows and columns. An ''n''-by-''n'' matrix is known as a square matrix of order Any two square matrices of the same order can be added and multiplied.
Square matrices are often ...
being viewed as the magnitude of the semiaxis of an -dimensional
ellipsoid
An ellipsoid is a surface that may be obtained from a sphere by deforming it by means of directional scalings, or more generally, of an affine transformation.
An ellipsoid is a quadric surface; that is, a surface that may be defined as the ...
. Similarly, the singular values of any matrix can be viewed as the magnitude of the semiaxis of an -dimensional
ellipsoid
An ellipsoid is a surface that may be obtained from a sphere by deforming it by means of directional scalings, or more generally, of an affine transformation.
An ellipsoid is a quadric surface; that is, a surface that may be defined as the ...
in -dimensional space, for example as an ellipse in a (tilted) 2D plane in a 3D space. Singular values encode magnitude of the semiaxis, while singular vectors encode direction. See
below for further details.
The columns of and are orthonormal bases
Since and are unitary, the columns of each of them form a set of
orthonormal vectors
In linear algebra, two vectors in an inner product space are orthonormal if they are orthogonal (or perpendicular along a line) unit vectors. A set of vectors form an orthonormal set if all vectors in the set are mutually orthogonal and all of un ...
, which can be regarded as
basis vectors
In mathematics, a set of vectors in a vector space is called a basis if every element of may be written in a unique way as a finite linear combination of elements of . The coefficients of this linear combination are referred to as componen ...
. The matrix maps the basis vector to the stretched unit vector . By the definition of a unitary matrix, the same is true for their conjugate transposes and , except the geometric interpretation of the singular values as stretches is lost. In short, the columns of , and are
orthonormal bases
In mathematics, particularly linear algebra, an orthonormal basis for an inner product space ''V'' with finite dimension is a basis for V whose vectors are orthonormal, that is, they are all unit vectors and orthogonal to each other. For example, ...
. When the
is a
positive-semidefinite Hermitian matrix
In mathematics, a Hermitian matrix (or self-adjoint matrix) is a complex square matrix that is equal to its own conjugate transpose—that is, the element in the -th row and -th column is equal to the complex conjugate of the element in the -th ...
, and are both equal to the unitary matrix used to diagonalize
. However, when
is not positive-semidefinite and Hermitian but still
diagonalizable
In linear algebra, a square matrix A is called diagonalizable or non-defective if it is similar to a diagonal matrix, i.e., if there exists an invertible matrix P and a diagonal matrix D such that or equivalently (Such D are not unique.) F ...
, its
eigendecomposition
In linear algebra, eigendecomposition is the factorization of a matrix into a canonical form, whereby the matrix is represented in terms of its eigenvalues and eigenvectors. Only diagonalizable matrices can be factorized in this way. When the matr ...
and singular value decomposition are distinct.
Geometric meaning
Because and are unitary, we know that the columns of yield an
orthonormal basis
In mathematics, particularly linear algebra, an orthonormal basis for an inner product space ''V'' with finite dimension is a basis for V whose vectors are orthonormal, that is, they are all unit vectors and orthogonal to each other. For example, ...
of and the columns of yield an orthonormal basis of (with respect to the standard
scalar product
In mathematics, the dot product or scalar productThe term ''scalar product'' means literally "product with a scalar as a result". It is also used sometimes for other symmetric bilinear forms, for example in a pseudo-Euclidean space. is an algebra ...
s on these spaces).
The
linear transformation
In mathematics, and more specifically in linear algebra, a linear map (also called a linear mapping, linear transformation, vector space homomorphism, or in some contexts linear function) is a mapping V \to W between two vector spaces that pre ...
:
has a particularly simple description with respect to these orthonormal bases: we have
:
where is the -th diagonal entry of
, and for .
The geometric content of the SVD theorem can thus be summarized as follows: for every linear map one can find orthonormal bases of and such that maps the -th basis vector of to a non-negative multiple of the -th basis vector of , and sends the left-over basis vectors to zero. With respect to these bases, the map is therefore represented by a diagonal matrix with non-negative real diagonal entries.
To get a more visual flavor of singular values and SVD factorization – at least when working on real vector spaces – consider the sphere of radius one in . The linear map maps this sphere onto an
ellipsoid
An ellipsoid is a surface that may be obtained from a sphere by deforming it by means of directional scalings, or more generally, of an affine transformation.
An ellipsoid is a quadric surface; that is, a surface that may be defined as the ...
in . Non-zero singular values are simply the lengths of the
semi-axes of this ellipsoid. Especially when , and all the singular values are distinct and non-zero, the SVD of the linear map can be easily analyzed as a succession of three consecutive moves: consider the ellipsoid and specifically its axes; then consider the directions in sent by onto these axes. These directions happen to be mutually orthogonal. Apply first an isometry sending these directions to the coordinate axes of . On a second move, apply an
endomorphism
In mathematics, an endomorphism is a morphism from a mathematical object to itself. An endomorphism that is also an isomorphism is an automorphism. For example, an endomorphism of a vector space is a linear map , and an endomorphism of a g ...
diagonalized along the coordinate axes and stretching or shrinking in each direction, using the semi-axes lengths of as stretching coefficients. The composition then sends the unit-sphere onto an ellipsoid isometric to . To define the third and last move, apply an isometry to this ellipsoid to obtain . As can be easily checked, the composition coincides with .
Example
Consider the matrix
:
A singular value decomposition of this matrix is given by
:
The scaling matrix
is zero outside of the diagonal (grey italics) and one diagonal element is zero (red bold). Furthermore, because the matrices and are
unitary
Unitary may refer to:
Mathematics
* Unitary divisor
* Unitary element
* Unitary group
* Unitary matrix
* Unitary morphism
* Unitary operator
* Unitary transformation
* Unitary representation
* Unitarity (physics)
* ''E''-unitary inverse semigrou ...
, multiplying by their respective conjugate transposes yields
identity matrices, as shown below. In this case, because and are real valued, each is an
orthogonal matrix
In linear algebra, an orthogonal matrix, or orthonormal matrix, is a real square matrix whose columns and rows are orthonormal vectors.
One way to express this is
Q^\mathrm Q = Q Q^\mathrm = I,
where is the transpose of and is the identity ma ...
.
:
This particular singular value decomposition is not unique. Choosing
such that
:
is also a valid singular value decomposition.
SVD and spectral decomposition
Singular values, singular vectors, and their relation to the SVD
A non-negative real number is a
singular value In mathematics, in particular functional analysis, the singular values, or ''s''-numbers of a compact operator T: X \rightarrow Y acting between Hilbert spaces X and Y, are the square roots of the (necessarily non-negative) eigenvalues of the self ...
for if and only if there exist unit-length vectors
in ''K
m'' and
in ''K
n'' such that
:
The vectors
and
are called left-singular and right-singular vectors for , respectively.
In any singular value decomposition
:
the diagonal entries of
are equal to the singular values of . The first columns of and are, respectively, left- and right-singular vectors for the corresponding singular values. Consequently, the above theorem implies that:
* An matrix has at most distinct singular values.
* It is always possible to find a
unitary basis for with a subset of basis vectors spanning the left-singular vectors of each singular value of .
* It is always possible to find a unitary basis for with a subset of basis vectors spanning the right-singular vectors of each singular value of .
A singular value for which we can find two left (or right) singular vectors that are linearly independent is called ''degenerate''. If
and
are two left-singular vectors which both correspond to the singular value σ, then any normalized linear combination of the two vectors is also a left-singular vector corresponding to the singular value σ. The similar statement is true for right-singular vectors. The number of independent left and right-singular vectors coincides, and these singular vectors appear in the same columns of and corresponding to diagonal elements of
all with the same value ''σ''.
As an exception, the left and right-singular vectors of singular value 0 comprise all unit vectors in the
cokernel
The cokernel of a linear mapping of vector spaces is the quotient space of the codomain of by the image of . The dimension of the cokernel is called the ''corank'' of .
Cokernels are dual to the kernels of category theory, hence the name: ...
and
kernel
Kernel may refer to:
Computing
* Kernel (operating system), the central component of most operating systems
* Kernel (image processing), a matrix used for image convolution
* Compute kernel, in GPGPU programming
* Kernel method, in machine learnin ...
, respectively, of , which by the
rank–nullity theorem
The rank–nullity theorem is a theorem in linear algebra, which asserts that the dimension of the domain of a linear map is the sum of its rank (the dimension of its image) and its ''nullity'' (the dimension of its kernel). p. 70, §2.1, Theor ...
cannot be the same dimension if . Even if all singular values are nonzero, if then the cokernel is nontrivial, in which case is padded with orthogonal vectors from the cokernel. Conversely, if , then is padded by orthogonal vectors from the kernel. However, if the singular value of 0 exists, the extra columns of or already appear as left or right-singular vectors.
Non-degenerate singular values always have unique left- and right-singular vectors, up to multiplication by a unit-phase factor ''e''
i''φ'' (for the real case up to a sign). Consequently, if all singular values of a square matrix are non-degenerate and non-zero, then its singular value decomposition is unique, up to multiplication of a column of by a unit-phase factor and simultaneous multiplication of the corresponding column of by the same unit-phase factor.
In general, the SVD is unique up to arbitrary unitary transformations applied uniformly to the column vectors of both and spanning the subspaces of each singular value, and up to arbitrary unitary transformations on vectors of and spanning the kernel and cokernel, respectively, of .
Relation to eigenvalue decomposition
The singular value decomposition is very general in the sense that it can be applied to any matrix, whereas
eigenvalue decomposition
In linear algebra, eigendecomposition is the factorization of a matrix into a canonical form, whereby the matrix is represented in terms of its eigenvalues and eigenvectors. Only diagonalizable matrices can be factorized in this way. When the matr ...
can only be applied to
diagonalizable matrices
In linear algebra, a square matrix A is called diagonalizable or non-defective if it is similar to a diagonal matrix, i.e., if there exists an invertible matrix P and a diagonal matrix D such that or equivalently (Such D are not unique.) ...
. Nevertheless, the two decompositions are related.
Given an SVD of , as described above, the following two relations hold:
:
The right-hand sides of these relations describe the eigenvalue decompositions of the left-hand sides. Consequently:
* The columns of (right-singular vectors) are
eigenvectors
In linear algebra, an eigenvector () or characteristic vector of a linear transformation is a nonzero vector that changes at most by a scalar factor when that linear transformation is applied to it. The corresponding eigenvalue, often denoted b ...
of .
* The columns of (left-singular vectors) are eigenvectors of .
* The non-zero elements of
(non-zero singular values) are the square roots of the non-zero
eigenvalues
In linear algebra, an eigenvector () or characteristic vector of a linear transformation is a nonzero vector that changes at most by a scalar factor when that linear transformation is applied to it. The corresponding eigenvalue, often denoted b ...
of or .
In the special case that is a
normal matrix In mathematics, a complex square matrix is normal if it commutes with its conjugate transpose :
The concept of normal matrices can be extended to normal operators on infinite dimensional normed spaces and to normal elements in C*-algebras. As ...
, which by definition must be square, the
spectral theorem
In mathematics, particularly linear algebra and functional analysis, a spectral theorem is a result about when a linear operator or matrix (mathematics), matrix can be Diagonalizable matrix, diagonalized (that is, represented as a diagonal matrix i ...
says that it can be
unitarily diagonalized
In linear algebra, a square matrix A is called diagonalizable or non-defective if it is similar to a diagonal matrix, i.e., if there exists an invertible matrix P and a diagonal matrix D such that or equivalently (Such D are not unique.) F ...
using a basis of
eigenvector
In linear algebra, an eigenvector () or characteristic vector of a linear transformation is a nonzero vector that changes at most by a scalar factor when that linear transformation is applied to it. The corresponding eigenvalue, often denoted b ...
s, so that it can be written for a unitary matrix and a diagonal matrix with complex elements along the diagonal. When is
positive semi-definite, will be non-negative real numbers so that the decomposition is also a singular value decomposition. Otherwise, it can be recast as an SVD by moving the phase of each to either its corresponding or . The natural connection of the SVD to non-normal matrices is through the
polar decomposition
In mathematics, the polar decomposition of a square real or complex matrix A is a factorization of the form A = U P, where U is an orthogonal matrix and P is a positive semi-definite symmetric matrix (U is a unitary matrix and P is a positive se ...
theorem: , where is positive semidefinite and normal, and is unitary.
Thus, except for positive semi-definite matrices, the eigenvalue decomposition and SVD of , while related, differ: the eigenvalue decomposition is , where is not necessarily unitary and is not necessarily positive semi-definite, while the SVD is , where
is diagonal and positive semi-definite, and and are unitary matrices that are not necessarily related except through the matrix . While only
non-defective square matrices have an eigenvalue decomposition, any
matrix has a SVD.
Applications of the SVD
Pseudoinverse
The singular value decomposition can be used for computing the
pseudoinverse
In mathematics, and in particular, algebra, a generalized inverse (or, g-inverse) of an element ''x'' is an element ''y'' that has some properties of an inverse element but not necessarily all of them. The purpose of constructing a generalized inv ...
of a matrix. (Various authors use different notation for the pseudoinverse; here we use .) Indeed, the pseudoinverse of the matrix with singular value decomposition is
:
where is the pseudoinverse of , which is formed by replacing every non-zero diagonal entry by its
reciprocal
Reciprocal may refer to:
In mathematics
* Multiplicative inverse, in mathematics, the number 1/''x'', which multiplied by ''x'' gives the product 1, also known as a ''reciprocal''
* Reciprocal polynomial, a polynomial obtained from another pol ...
and transposing the resulting matrix. The pseudoinverse is one way to solve
linear least squares
Linear least squares (LLS) is the least squares approximation of linear functions to data.
It is a set of formulations for solving statistical problems involved in linear regression, including variants for ordinary (unweighted), weighted, and ...
problems.
Solving homogeneous linear equations
A set of
homogeneous linear equation
In mathematics, a system of linear equations (or linear system) is a collection of one or more linear equations involving the same variables.
For example,
:\begin
3x+2y-z=1\\
2x-2y+4z=-2\\
-x+\fracy-z=0
\end
is a system of three equations in th ...
s can be written as for a matrix and vector . A typical situation is that is known and a non-zero is to be determined which satisfies the equation. Such an belongs to 's
null space
In mathematics, the kernel of a linear map, also known as the null space or nullspace, is the linear subspace of the Domain of a function, domain of the map which is mapped to the zero vector. That is, given a linear map between two vector space ...
and is sometimes called a (right) null vector of . The vector can be characterized as a right-singular vector corresponding to a singular value of that is zero. This observation means that if is a
square matrix
In mathematics, a square matrix is a matrix with the same number of rows and columns. An ''n''-by-''n'' matrix is known as a square matrix of order Any two square matrices of the same order can be added and multiplied.
Square matrices are often ...
and has no vanishing singular value, the equation has no non-zero as a solution. It also means that if there are several vanishing singular values, any linear combination of the corresponding right-singular vectors is a valid solution. Analogously to the definition of a (right) null vector, a non-zero satisfying , with denoting the conjugate transpose of , is called a left null vector of .
Total least squares minimization
A
total least squares
In applied statistics, total least squares is a type of errors-in-variables regression, a least squares data modeling technique in which observational errors on both dependent and independent variables are taken into account. It is a generaliza ...
problem seeks the vector that minimizes the
2-norm of a vector under the constraint . The solution turns out to be the right-singular vector of corresponding to the smallest singular value.
Range, null space and rank
Another application of the SVD is that it provides an explicit representation of the
range
Range may refer to:
Geography
* Range (geographic), a chain of hills or mountains; a somewhat linear, complex mountainous or hilly area (cordillera, sierra)
** Mountain range, a group of mountains bordered by lowlands
* Range, a term used to i ...
and
null space
In mathematics, the kernel of a linear map, also known as the null space or nullspace, is the linear subspace of the Domain of a function, domain of the map which is mapped to the zero vector. That is, given a linear map between two vector space ...
of a matrix . The right-singular vectors corresponding to vanishing singular values of span the null space of and the left-singular vectors corresponding to the non-zero singular values of span the range of . For example, in the above
example
Example may refer to:
* '' exempli gratia'' (e.g.), usually read out in English as "for example"
* .example, reserved as a domain name that may not be installed as a top-level domain of the Internet
** example.com, example.net, example.org, ex ...
the null space is spanned by the last two rows of and the range is spanned by the first three columns of .
As a consequence, the
rank
Rank is the relative position, value, worth, complexity, power, importance, authority, level, etc. of a person or object within a ranking, such as:
Level or position in a hierarchical organization
* Academic rank
* Diplomatic rank
* Hierarchy
* H ...
of equals the number of non-zero singular values which is the same as the number of non-zero diagonal elements in
. In numerical linear algebra the singular values can be used to determine the ''effective rank'' of a matrix, as
rounding error
A roundoff error, also called rounding error, is the difference between the result produced by a given algorithm using exact arithmetic and the result produced by the same algorithm using finite-precision, rounded arithmetic. Rounding errors are d ...
may lead to small but non-zero singular values in a rank deficient matrix. Singular values beyond a significant gap are assumed to be numerically equivalent to zero.
Low-rank matrix approximation
Some practical applications need to solve the problem of
approximating a matrix with another matrix
, said to be
truncated, which has a specific rank . In the case that the approximation is based on minimizing the
Frobenius norm
In mathematics, a matrix norm is a vector norm in a vector space whose elements (vectors) are matrices (of given dimensions).
Preliminaries
Given a field K of either real or complex numbers, let K^ be the -vector space of matrices with m rows ...
of the difference between and
under the constraint that
, it turns out that the solution is given by the SVD of , namely
:
where
is the same matrix as
except that it contains only the largest singular values (the other singular values are replaced by zero). This is known as the
Eckart–Young theorem
In linear algebra, the singular value decomposition (SVD) is a factorization of a real or complex matrix. It generalizes the eigendecomposition of a square normal matrix with an orthonormal eigenbasis to any \ m \times n\ matrix. It is relate ...
, as it was proved by those two authors in 1936 (although it was later found to have been known to earlier authors; see ).
Separable models
The SVD can be thought of as decomposing a matrix into a weighted, ordered sum of separable matrices. By separable, we mean that a matrix can be written as an
outer product
In linear algebra, the outer product of two coordinate vector
In linear algebra, a coordinate vector is a representation of a vector as an ordered list of numbers (a tuple) that describes the vector in terms of a particular ordered basis. An ea ...
of two vectors , or, in coordinates,
. Specifically, the matrix can be decomposed as
:
Here and are the -th columns of the corresponding SVD matrices, are the ordered singular values, and each is separable. The SVD can be used to find the decomposition of an image processing filter into separable horizontal and vertical filters. Note that the number of non-zero is exactly the rank of the matrix.
Separable models often arise in biological systems, and the SVD factorization is useful to analyze such systems. For example, some visual area V1 simple cells' receptive fields can be well described by a
Gabor filter
In image processing, a Gabor filter, named after Dennis Gabor, is a linear filter used for texture analysis, which essentially means that it analyzes whether there is any specific frequency content in the image in specific directions in a localiz ...
in the space domain multiplied by a modulation function in the time domain. Thus, given a linear filter evaluated through, for example,
reverse correlation, one can rearrange the two spatial dimensions into one dimension, thus yielding a two-dimensional filter (space, time) which can be decomposed through SVD. The first column of in the SVD factorization is then a Gabor while the first column of represents the time modulation (or vice versa). One may then define an index of separability
:
which is the fraction of the power in the matrix M which is accounted for by the first separable matrix in the decomposition.
Nearest orthogonal matrix
It is possible to use the SVD of a square matrix to determine the
orthogonal matrix
In linear algebra, an orthogonal matrix, or orthonormal matrix, is a real square matrix whose columns and rows are orthonormal vectors.
One way to express this is
Q^\mathrm Q = Q Q^\mathrm = I,
where is the transpose of and is the identity ma ...
closest to . The closeness of fit is measured by the
Frobenius norm
In mathematics, a matrix norm is a vector norm in a vector space whose elements (vectors) are matrices (of given dimensions).
Preliminaries
Given a field K of either real or complex numbers, let K^ be the -vector space of matrices with m rows ...
of . The solution is the product . This intuitively makes sense because an orthogonal matrix would have the decomposition where is the identity matrix, so that if then the product amounts to replacing the singular values with ones. Equivalently, the solution is the unitary matrix of the Polar Decomposition in either order of stretch and rotation, as described above.
A similar problem, with interesting applications in
shape analysis, is the
orthogonal Procrustes problem The orthogonal Procrustes problem is a matrix approximation problem in linear algebra. In its classical form, one is given two matrices A and B and asked to find an orthogonal matrix \Omega which most closely maps A to B. Specifically,
:R = \arg\m ...
, which consists of finding an orthogonal matrix which most closely maps to . Specifically,
:
where
denotes the Frobenius norm.
This problem is equivalent to finding the nearest orthogonal matrix to a given matrix .
The Kabsch algorithm
The
Kabsch algorithm The Kabsch algorithm, named after Wolfgang Kabsch, is a method for calculating the optimal rotation matrix that minimizes the RMSD ( root mean squared deviation) between two paired sets of points. It is useful in graphics, cheminformatics to compar ...
(called
Wahba's problem
In applied mathematics, Wahba's problem, first posed by Grace Wahba in 1965, seeks to find a rotation matrix ( special orthogonal matrix) between two coordinate systems from a set of (weighted) vector observations. Solutions to Wahba's problem are ...
in other fields) uses SVD to compute the optimal rotation (with respect to least-squares minimization) that will align a set of points with a corresponding set of points. It is used, among other applications, to compare the structures of molecules.
Signal processing
The SVD and pseudoinverse have been successfully applied to
signal processing
Signal processing is an electrical engineering subfield that focuses on analyzing, modifying and synthesizing ''signals'', such as audio signal processing, sound, image processing, images, and scientific measurements. Signal processing techniq ...
,
image processing
An image is a visual representation of something. It can be two-dimensional, three-dimensional, or somehow otherwise feed into the visual system to convey information. An image can be an artifact, such as a photograph or other two-dimensiona ...
and
big data
Though used sometimes loosely partly because of a lack of formal definition, the interpretation that seems to best describe Big data is the one associated with large body of information that we could not comprehend when used only in smaller am ...
(e.g., in genomic signal processing).
Other examples
The SVD is also applied extensively to the study of linear
inverse problem
An inverse problem in science is the process of calculating from a set of observations the causal factors that produced them: for example, calculating an image in X-ray computed tomography, source reconstruction in acoustics, or calculating the ...
s and is useful in the analysis of regularization methods such as that of
Tikhonov. It is widely used in statistics, where it is related to
principal component analysis
Principal component analysis (PCA) is a popular technique for analyzing large datasets containing a high number of dimensions/features per observation, increasing the interpretability of data while preserving the maximum amount of information, and ...
and to
correspondence analysis, and in
signal processing
Signal processing is an electrical engineering subfield that focuses on analyzing, modifying and synthesizing ''signals'', such as audio signal processing, sound, image processing, images, and scientific measurements. Signal processing techniq ...
and
pattern recognition
Pattern recognition is the automated recognition of patterns and regularities in data. It has applications in statistical data analysis, signal processing, image analysis, information retrieval, bioinformatics, data compression, computer graphi ...
. It is also used in output-only
modal analysis
Modal analysis is the study of the dynamic properties of systems in the frequency domain. Examples would include measuring the vibration of a car's body when it is attached to a shaker, or the noise pattern in a room when excited by a loudspeak ...
, where the non-scaled
mode shape
A normal mode of a dynamical system is a pattern of motion in which all parts of the system move sinusoidally with the same frequency and with a fixed phase relation. The free motion described by the normal modes takes place at fixed frequencies. ...
s can be determined from the singular vectors. Yet another usage is
latent semantic indexing
Latent semantic analysis (LSA) is a technique in natural language processing, in particular distributional semantics, of analyzing relationships between a set of documents and the terms they contain by producing a set of concepts related to the do ...
in natural-language text processing.
In general numerical computation involving linear or linearized systems, there is a universal constant that characterizes the regularity or singularity of a problem, which is the system's "condition number"
. It often controls the error rate or convergence rate of a given computational scheme on such systems.
The SVD also plays a crucial role in the field of
quantum information
Quantum information is the information of the state of a quantum system. It is the basic entity of study in quantum information theory, and can be manipulated using quantum information processing techniques. Quantum information refers to both th ...
, in a form often referred to as the
Schmidt decomposition
In linear algebra, the Schmidt decomposition (named after its originator Erhard Schmidt) refers to a particular way of expressing a vector in the tensor product of two inner product spaces. It has numerous applications in quantum information the ...
. Through it, states of two quantum systems are naturally decomposed, providing a necessary and sufficient condition for them to be
entangled: if the rank of the
matrix is larger than one.
One application of SVD to rather large matrices is in
numerical weather prediction
Numerical weather prediction (NWP) uses mathematical models of the atmosphere and oceans to predict the weather based on current weather conditions. Though first attempted in the 1920s, it was not until the advent of computer simulation in th ...
, where
Lanczos methods are used to estimate the most linearly quickly growing few perturbations to the central numerical weather prediction over a given initial forward time period; i.e., the singular vectors corresponding to the largest singular values of the linearized propagator for the global weather over that time interval. The output singular vectors in this case are entire weather systems. These perturbations are then run through the full nonlinear model to generate an
ensemble forecast, giving a handle on some of the uncertainty that should be allowed for around the current central prediction.
SVD has also been applied to reduced order modelling. The aim of reduced order modelling is to reduce the number of degrees of freedom in a complex system which is to be modeled. SVD was coupled with
radial basis functions 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\, ), or some other fixe ...
to interpolate solutions to three-dimensional unsteady flow problems.
Interestingly, SVD has been used to improve gravitational waveform modeling by the ground-based gravitational-wave interferometer aLIGO. SVD can help to increase the accuracy and speed of waveform generation to support gravitational-waves searches and update two different waveform models.
Singular value decomposition is used in
recommender systems
A recommender system, or a recommendation system (sometimes replacing 'system' with a synonym such as platform or engine), is a subclass of information filtering system that provide suggestions for items that are most pertinent to a particular u ...
to predict people's item ratings. Distributed algorithms have been developed for the purpose of calculating the SVD on clusters of commodity machines.
Low-rank SVD has been applied for hotspot detection from spatiotemporal data with application to disease
outbreak
In epidemiology, an outbreak is a sudden increase in occurrences of a disease when cases are in excess of normal expectancy for the location or season. It may affect a small and localized group or impact upon thousands of people across an entire ...
detection. A combination of SVD and
higher-order SVD also has been applied for real time event detection from complex data streams (multivariate data with space and time dimensions) in
disease surveillance
Disease surveillance is an epidemiological practice by which the spread of disease is monitored in order to establish patterns of progression. The main role of disease surveillance is to predict, observe, and minimize the harm caused by outbreak, ...
.
Existence proofs
An eigenvalue of a matrix is characterized by the algebraic relation . When is
Hermitian {{Short description, none
Numerous things are named after the French mathematician Charles Hermite (1822–1901):
Hermite
* Cubic Hermite spline, a type of third-degree spline
* Gauss–Hermite quadrature, an extension of Gaussian quadrature m ...
, a variational characterization is also available. Let be a real
symmetric matrix
In linear algebra, a symmetric matrix is a square matrix that is equal to its transpose. Formally,
Because equal matrices have equal dimensions, only square matrices can be symmetric.
The entries of a symmetric matrix are symmetric with re ...
. Define
:
By the
extreme value theorem
In calculus, the extreme value theorem states that if a real-valued function f is continuous on the closed interval ,b/math>, then f must attain a maximum and a minimum, each at least once. That is, there exist numbers c and d in ,b/math> suc ...
, this continuous function attains a maximum at some u when restricted to the unit sphere {, , x, , = 1}. By the
Lagrange multipliers
In mathematical optimization, the method of Lagrange multipliers is a strategy for finding the local maxima and minima of a function subject to equality constraints (i.e., subject to the condition that one or more equations have to be satisfied ex ...
theorem, u necessarily satisfies
:
for some real number . The nabla symbol, , is the
del
Del, or nabla, is an operator used in mathematics (particularly in vector calculus) as a vector differential operator, usually represented by the nabla symbol ∇. When applied to a function defined on a one-dimensional domain, it denotes ...
operator (differentiation with respect to x). Using the symmetry of we obtain
:
Therefore , so u is a unit length eigenvector of . For every unit length eigenvector v of its eigenvalue is ''f''(v), so is the largest eigenvalue of . The same calculation performed on the orthogonal complement of ''u'' gives the next largest eigenvalue and so on. The complex Hermitian case is similar; there is a real-valued function of real variables.
Singular values are similar in that they can be described algebraically or from variational principles. Although, unlike the eigenvalue case, Hermiticity, or symmetry, of is no longer required.
This section gives these two arguments for existence of singular value decomposition.
Based on the spectral theorem
Let
be an complex matrix. Since
is positive semi-definite and Hermitian, by the
spectral theorem
In mathematics, particularly linear algebra and functional analysis, a spectral theorem is a result about when a linear operator or matrix (mathematics), matrix can be Diagonalizable matrix, diagonalized (that is, represented as a diagonal matrix i ...
, there exists an unitary matrix
such that
:
where
is diagonal and positive definite, of dimension
, with
the number of non-zero eigenvalues of
(which can be shown to verify
). Note that
is here by definition a matrix whose
-th column is the
-th eigenvector of
, corresponding to the eigenvalue
. Moreover, the
-th column of
, for
, is an eigenvector of
with eigenvalue
. This can be expressed by writing
as
, where the columns of
and
therefore contain the eigenvectors of
corresponding to non-zero and zero eigenvalues, respectively. Using this rewriting of
, the equation becomes:
:
This implies that
:
Moreover, the second equation implies
. Finally, the unitary-ness of
translates, in terms of
and
, into the following conditions:
:
where the subscripts on the identity matrices are used to remark that they are of different dimensions.
Let us now define
:
Then,
:
since
This can be also seen as immediate consequence of the fact that
. This is equivalent to the observation that if
is the set of eigenvectors of
corresponding to non-vanishing eigenvalues
, then
is a set of orthogonal vectors, and
is a (generally not complete) set of ''orthonormal'' vectors. This matches with the matrix formalism used above denoting with
the matrix whose columns are
, with
the matrix whose columns are the eigenvectors of
with vanishing eigenvalue, and
the matrix whose columns are the vectors
.
We see that this is almost the desired result, except that
and
are in general not unitary, since they might not be square. However, we do know that the number of rows of
is no smaller than the number of columns, since the dimensions of
is no greater than
and
. Also, since
:
the columns in
are orthonormal and can be extended to an orthonormal basis. This means that we can choose
such that
is unitary.
For we already have to make it unitary. Now, define
:
where extra zero rows are added or removed to make the number of zero rows equal the number of columns of , and hence the overall dimensions of
equal to
. Then
:
which is the desired result:
:
Notice the argument could begin with diagonalizing rather than (This shows directly that and have the same non-zero eigenvalues).
Based on variational characterization
The singular values can also be characterized as the maxima of , considered as a function of and , over particular subspaces. The singular vectors are the values of and where these maxima are attained.
Let denote an matrix with real entries. Let be the unit
-sphere in
, and define
Consider the function restricted to . Since both and are
compact
Compact as used in politics may refer broadly to a pact or treaty; in more specific cases it may refer to:
* Interstate compact
* Blood compact, an ancient ritual of the Philippines
* Compact government, a type of colonial rule utilized in British ...
sets, their
product
Product may refer to:
Business
* Product (business), an item that serves as a solution to a specific consumer problem.
* Product (project management), a deliverable or set of deliverables that contribute to a business solution
Mathematics
* Produ ...
is also compact. Furthermore, since is continuous, it attains a largest value for at least one pair of vectors and . This largest value is denoted and the corresponding vectors are denoted and . Since is the largest value of it must be non-negative. If it were negative, changing the sign of either or would make it positive and therefore larger.
Statement. are left and right-singular vectors of with corresponding singular value ''σ''
1.
Proof. Similar to the eigenvalues case, by assumption the two vectors satisfy the Lagrange multiplier equation:
:
After some algebra, this becomes
:
Multiplying the first equation from left by
and the second equation from left by
and taking into account gives
:
Plugging this into the pair of equations above, we have
:
This proves the statement.
More singular vectors and singular values can be found by maximizing over normalized which are orthogonal to and , respectively.
The passage from real to complex is similar to the eigenvalue case.
Calculating the SVD
The singular value decomposition can be computed using the following observations:
* The left-singular vectors of are a set of
orthonormal
In linear algebra, two vectors in an inner product space are orthonormal if they are orthogonal (or perpendicular along a line) unit vectors. A set of vectors form an orthonormal set if all vectors in the set are mutually orthogonal and all of un ...
eigenvectors
In linear algebra, an eigenvector () or characteristic vector of a linear transformation is a nonzero vector that changes at most by a scalar factor when that linear transformation is applied to it. The corresponding eigenvalue, often denoted b ...
of .
* The right-singular vectors of are a set of orthonormal eigenvectors of .
* The non-zero singular values of (found on the diagonal entries of
) are the square roots of the non-zero
eigenvalues
In linear algebra, an eigenvector () or characteristic vector of a linear transformation is a nonzero vector that changes at most by a scalar factor when that linear transformation is applied to it. The corresponding eigenvalue, often denoted b ...
of both and .
Numerical approach
The SVD of a matrix is typically computed by a two-step procedure. In the first step, the matrix is reduced to a
bidiagonal matrix. This takes
O(''mn''
2) floating-point operations (flop), assuming that ''m'' ≥ ''n''. The second step is to compute the SVD of the bidiagonal matrix. This step can only be done with an
iterative method
In computational mathematics, an iterative method is a Algorithm, mathematical procedure that uses an initial value to generate a sequence of improving approximate solutions for a class of problems, in which the ''n''-th approximation is derived fr ...
(as with
eigenvalue algorithm
In numerical analysis, one of the most important problems is designing efficient and stable algorithms for finding the eigenvalues of a matrix. These eigenvalue algorithms may also find eigenvectors.
Eigenvalues and eigenvectors
Given an square ...
s). However, in practice it suffices to compute the SVD up to a certain precision, like the
machine epsilon
Machine epsilon or machine precision is an upper bound on the relative approximation error due to rounding in floating point arithmetic. This value characterizes computer arithmetic in the field of numerical analysis, and by extension in the subjec ...
. If this precision is considered constant, then the second step takes O(''n'') iterations, each costing O(''n'') flops. Thus, the first step is more expensive, and the overall cost is O(''mn''
2) flops .
The first step can be done using
Householder reflection
In linear algebra, a Householder transformation (also known as a Householder reflection or elementary reflector) is a linear transformation that describes a reflection about a plane or hyperplane containing the origin. The Householder transformati ...
s for a cost of 4''mn''
2 − 4''n''
3/3 flops, assuming that only the singular values are needed and not the singular vectors. If ''m'' is much larger than ''n'' then it is advantageous to first reduce the matrix to a triangular matrix with the
QR decomposition
In linear algebra, a QR decomposition, also known as a QR factorization or QU factorization, is a decomposition of a matrix ''A'' into a product ''A'' = ''QR'' of an orthogonal matrix ''Q'' and an upper triangular matrix ''R''. QR decompo ...
and then use Householder reflections to further reduce the matrix to bidiagonal form; the combined cost is 2''mn''
2 + 2''n''
3 flops .
The second step can be done by a variant of the
QR algorithm
In numerical linear algebra, the QR algorithm or QR iteration is an eigenvalue algorithm: that is, a procedure to calculate the eigenvalues and eigenvectors of a matrix. The QR algorithm was developed in the late 1950s by John G. F. Francis and by ...
for the computation of eigenvalues, which was first described by . The
LAPACK
LAPACK ("Linear Algebra Package") is a standard software library for numerical linear algebra. It provides routines for solving systems of linear equations and linear least squares, eigenvalue problems, and singular value decomposition. It als ...
subroutine DBDSQR implements this iterative method, with some modifications to cover the case where the singular values are very small . Together with a first step using Householder reflections and, if appropriate, QR decomposition, this forms the DGESVD routine for the computation of the singular value decomposition.
The same algorithm is implemented in the
GNU Scientific Library
The GNU Scientific Library (or GSL) is a software library for numerical computations in applied mathematics and science. The GSL is written in C; wrappers are available for other programming languages. The GSL is part of the GNU Project and is d ...
(GSL). The GSL also offers an alternative method that uses a one-sided
Jacobi orthogonalization Jacobi may refer to:
* People with the surname Jacobi
Mathematics:
* Jacobi sum, a type of character sum
* Jacobi method, a method for determining the solutions of a diagonally dominant system of linear equations
* Jacobi eigenvalue algorithm, a ...
in step 2 . This method computes the SVD of the bidiagonal matrix by solving a sequence of 2 × 2 SVD problems, similar to how the
Jacobi eigenvalue algorithm In numerical linear algebra, the Jacobi eigenvalue algorithm is an iterative method for the calculation of the eigenvalues and eigenvectors of a real symmetric matrix (a process known as diagonalization). It is named after Carl Gustav Jacob Jacobi, ...
solves a sequence of 2 × 2 eigenvalue methods . Yet another method for step 2 uses the idea of
divide-and-conquer eigenvalue algorithm Divide-and-conquer eigenvalue algorithms are a class of eigenvalue algorithms for Hermitian or real symmetric matrices that have recently (circa 1990s) become competitive in terms of stability and efficiency with more traditional algorithms such as ...
s .
There is an alternative way that does not explicitly use the eigenvalue decomposition. Usually the singular value problem of a matrix is converted into an equivalent symmetric eigenvalue problem such as , , or
:
The approaches that use eigenvalue decompositions are based on the
QR algorithm
In numerical linear algebra, the QR algorithm or QR iteration is an eigenvalue algorithm: that is, a procedure to calculate the eigenvalues and eigenvectors of a matrix. The QR algorithm was developed in the late 1950s by John G. F. Francis and by ...
, which is well-developed to be stable and fast.
Note that the singular values are real and right- and left- singular vectors are not required to form similarity transformations. One can iteratively alternate between the
QR decomposition
In linear algebra, a QR decomposition, also known as a QR factorization or QU factorization, is a decomposition of a matrix ''A'' into a product ''A'' = ''QR'' of an orthogonal matrix ''Q'' and an upper triangular matrix ''R''. QR decompo ...
and the
LQ decomposition
In linear algebra, a QR decomposition, also known as a QR factorization or QU factorization, is a matrix decomposition, decomposition of a matrix (mathematics), matrix ''A'' into a product ''A'' = ''QR'' of an orthogonal matrix ''Q'' and ...
to find the real diagonal
Hermitian matrices
In mathematics, a Hermitian matrix (or self-adjoint matrix) is a complex square matrix that is equal to its own conjugate transpose—that is, the element in the -th row and -th column is equal to the complex conjugate of the element in the -th ...
. The
QR decomposition
In linear algebra, a QR decomposition, also known as a QR factorization or QU factorization, is a decomposition of a matrix ''A'' into a product ''A'' = ''QR'' of an orthogonal matrix ''Q'' and an upper triangular matrix ''R''. QR decompo ...
gives and the
LQ decomposition
In linear algebra, a QR decomposition, also known as a QR factorization or QU factorization, is a matrix decomposition, decomposition of a matrix (mathematics), matrix ''A'' into a product ''A'' = ''QR'' of an orthogonal matrix ''Q'' and ...
of gives . Thus, at every iteration, we have , update and repeat the orthogonalizations. Eventually, this iteration between
QR decomposition
In linear algebra, a QR decomposition, also known as a QR factorization or QU factorization, is a decomposition of a matrix ''A'' into a product ''A'' = ''QR'' of an orthogonal matrix ''Q'' and an upper triangular matrix ''R''. QR decompo ...
and
LQ decomposition
In linear algebra, a QR decomposition, also known as a QR factorization or QU factorization, is a matrix decomposition, decomposition of a matrix (mathematics), matrix ''A'' into a product ''A'' = ''QR'' of an orthogonal matrix ''Q'' and ...
produces left- and right- unitary singular matrices. This approach cannot readily be accelerated, as the QR algorithm can with spectral shifts or deflation. This is because the shift method is not easily defined without using similarity transformations. However, this iterative approach is very simple to implement, so is a good choice when speed does not matter. This method also provides insight into how purely orthogonal/unitary transformations can obtain the SVD.
Analytic result of 2 × 2 SVD
The singular values of a 2 × 2 matrix can be found analytically. Let the matrix be
where
are complex numbers that parameterize the matrix, is the identity matrix, and
denote the
Pauli matrices
In mathematical physics and mathematics, the Pauli matrices are a set of three complex matrices which are Hermitian, involutory and unitary. Usually indicated by the Greek letter sigma (), they are occasionally denoted by tau () when used in ...
. Then its two singular values are given by
:
Reduced SVDs
In applications it is quite unusual for the full SVD, including a full unitary decomposition of the null-space of the matrix, to be required. Instead, it is often sufficient (as well as faster, and more economical for storage) to compute a reduced version of the SVD. The following can be distinguished for an ''m''×''n'' matrix ''M'' of rank ''r'':
Thin SVD
The thin, or economy-sized, SVD of a matrix ''M'' is given by
:
where
:
,
the matrices ''U''
''k'' and ''V''
''k'' contain only the first ''k'' columns of ''U'' and ''V'', and Σ
''k'' contains only the first ''k'' singular values from Σ. The matrix ''U''
''k'' is thus ''m''×''k'', Σ
''k'' is ''k''×''k'' diagonal, and ''V''
''k''* is ''k''×''n''.
The thin SVD uses significantly less space and computation time if ''k'' ≪ max(''m'', ''n''). The first stage in its calculation will usually be a
QR decomposition
In linear algebra, a QR decomposition, also known as a QR factorization or QU factorization, is a decomposition of a matrix ''A'' into a product ''A'' = ''QR'' of an orthogonal matrix ''Q'' and an upper triangular matrix ''R''. QR decompo ...
of ''M'', which can make for a significantly quicker calculation in this case.
Compact SVD
:
Only the ''r'' column vectors of ''U'' and ''r'' row vectors of ''V*'' corresponding to the non-zero singular values Σ
''r'' are calculated. The remaining vectors of ''U'' and ''V*'' are not calculated. This is quicker and more economical than the thin SVD if ''r'' ≪ min(''m'', ''n''). The matrix ''U''
''r'' is thus ''m''×''r'', Σ
''r'' is ''r''×''r'' diagonal, and ''V''
''r''* is ''r''×''n''.
Truncated SVD
In many applications the number ''r'' of the non-zero singular values is large making even the Compact SVD impractical to compute. In such cases, the smallest singular values may need to be truncated to compute only ''t'' ≪ ''r'' non-zero singular values. The truncated SVD is no longer an exact decomposition of the original matrix ''M'', but rather provides the optimal
low-rank matrix approximation by any matrix of a fixed rank ''t''
:
,
where matrix ''U''
''t'' is ''m''×''t'', Σ
''t'' is ''t''×''t'' diagonal, and ''V''
''t''* is ''t''×''n''.
Only the ''t'' column vectors of ''U'' and ''t'' row vectors of ''V*'' corresponding to the ''t'' largest singular values Σ
''t'' are calculated. This can be much quicker and more economical than the compact SVD if ''t''≪''r'', but requires a completely different toolset of numerical solvers.
In applications that require an approximation to the
Moore–Penrose inverse
In mathematics, and in particular linear algebra, the Moore–Penrose inverse of a matrix is the most widely known generalization of the inverse matrix. It was independently described by E. H. Moore in 1920, Arne Bjerhammar in 1951, and Roger Pe ...
of the matrix ''M'', the smallest singular values of ''M'' are of interest, which are more challenging to compute compared to the largest ones.
Truncated SVD is employed in
latent semantic indexing
Latent semantic analysis (LSA) is a technique in natural language processing, in particular distributional semantics, of analyzing relationships between a set of documents and the terms they contain by producing a set of concepts related to the do ...
.
Norms
Ky Fan norms
The sum of the ''k'' largest singular values of ''M'' is a
matrix norm
In mathematics, a matrix norm is a vector norm in a vector space whose elements (vectors) are matrices (of given dimensions).
Preliminaries
Given a field K of either real or complex numbers, let K^ be the -vector space of matrices with m rows ...
, the
Ky Fan
Ky Fan (樊𰋀, , September 19, 1914 – March 22, 2010) was a Chinese-born American mathematician. He was a professor of mathematics at the University of California, Santa Barbara.
Biography
Fan was born in Hangzhou, the capital of Zhejian ...
''k''-norm of ''M''.
The first of the Ky Fan norms, the Ky Fan 1-norm, is the same as the
operator norm
In mathematics, the operator norm measures the "size" of certain linear operators by assigning each a real number called its . Formally, it is a norm defined on the space of bounded linear operators between two given normed vector spaces.
Introdu ...
of ''M'' as a linear operator with respect to the Euclidean norms of ''K''
''m'' and ''K''
''n''. In other words, the Ky Fan 1-norm is the operator norm induced by the standard ''ℓ''
2 Euclidean inner product. For this reason, it is also called the operator 2-norm. One can easily verify the relationship between the Ky Fan 1-norm and singular values. It is true in general, for a bounded operator ''M'' on (possibly infinite-dimensional) Hilbert spaces
:
But, in the matrix case, (''M* M'')
1/2 is a
normal matrix In mathematics, a complex square matrix is normal if it commutes with its conjugate transpose :
The concept of normal matrices can be extended to normal operators on infinite dimensional normed spaces and to normal elements in C*-algebras. As ...
, so , , ''M* M'', ,
1/2 is the largest eigenvalue of (''M* M'')
1/2, i.e. the largest singular value of ''M''.
The last of the Ky Fan norms, the sum of all singular values, is the
trace norm (also known as the 'nuclear norm'), defined by , , ''M'', , = Tr
1/2">''M* M'')1/2(the eigenvalues of ''M* M'' are the squares of the singular values).
Hilbert–Schmidt norm
The singular values are related to another norm on the space of operators. Consider the
Hilbert–Schmidt inner product on the matrices, defined by
:
So the induced norm is
:
Since the trace is invariant under unitary equivalence, this shows
:
where are the singular values of . This is called the
Frobenius norm
In mathematics, a matrix norm is a vector norm in a vector space whose elements (vectors) are matrices (of given dimensions).
Preliminaries
Given a field K of either real or complex numbers, let K^ be the -vector space of matrices with m rows ...
, Schatten 2-norm, or Hilbert–Schmidt norm of . Direct calculation shows that the Frobenius norm of coincides with:
:
In addition, the Frobenius norm and the trace norm (the nuclear norm) are special cases of the
Schatten norm
In mathematics, specifically functional analysis, the Schatten norm (or Schatten–von-Neumann norm)
arises as a generalization of ''p''-integrability similar to the trace class norm and the Hilbert–Schmidt norm.
Definition
Let H_1, H_2 be H ...
.
Variations and generalizations
Mode-''k'' representation
can be represented using
mode-''k'' multiplication of matrix
applying
then
on the result; that is
.
Tensor SVD
Two types of tensor decompositions exist, which generalise the SVD to multi-way arrays. One of them decomposes a tensor into a sum of rank-1 tensors, which is called a
tensor rank decomposition. The second type of decomposition computes the orthonormal subspaces associated with the different factors appearing in the tensor product of vector spaces in which the tensor lives. This decomposition is referred to in the literature as the
higher-order SVD (HOSVD) or
Tucker3/TuckerM. In addition,
multilinear principal component analysis Within statistics, Multilinear principal component analysis (MPCA) is a multilinear extension of principal component analysis (PCA). MPCA is employed in the analysis of M-way arrays, i.e. a cube or hyper-cube of numbers, also informally referred t ...
in
multilinear subspace learning
Multilinear subspace learning is an approach to dimensionality reduction.M. A. O. Vasilescu, D. Terzopoulos (2003"Multilinear Subspace Analysis of Image Ensembles" "Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVP ...
involves the same mathematical operations as Tucker decomposition, being used in a different context of
dimensionality reduction
Dimensionality reduction, or dimension reduction, is the transformation of data from a high-dimensional space into a low-dimensional space so that the low-dimensional representation retains some meaningful properties of the original data, ideally ...
.
Scale-invariant SVD
The singular values of a matrix ''A'' are uniquely defined and are invariant with respect to left and/or right unitary transformations of ''A''. In other words, the singular values of ''UAV'', for unitary ''U'' and ''V'', are equal to the singular values of ''A''. This is an important property for applications in which it is necessary to preserve Euclidean distances and invariance with respect to rotations.
The Scale-Invariant SVD, or SI-SVD, is analogous to the conventional SVD except that its uniquely-determined singular values are invariant with respect to diagonal transformations of ''A''. In other words, the singular values of ''DAE'', for invertible diagonal matrices ''D'' and ''E'', are equal to the singular values of ''A''. This is an important property for applications for which invariance to the choice of units on variables (e.g., metric versus imperial units) is needed.
Higher-order SVD of functions (HOSVD)
Tensor product (TP) model transformation numerically reconstruct the
HOSVD
In multilinear algebra, the higher-order singular value decomposition (HOSVD) of a tensor is a specific orthogonal Tucker decomposition. It may be regarded as one generalization of the matrix singular value decomposition. It has applications in co ...
of functions. For further details please visit:
*
Tensor product model transformation In mathematics, the tensor product (TP) model transformation was proposed by Baranyi and Yam as key concept for higher-order singular value decomposition of functions. It transforms a function (which can be given via closed formulas or neural netw ...
*
HOSVD-based canonical form of TP functions and qLPV models Based on the key idea of higher-order singular value decomposition (HOSVD) in tensor algebra, Baranyi and Yam proposed the concept of HOSVD-based canonical form of TP functions and quasi-LPV system models. Szeidl et al. proved that the TP model tran ...
*
TP model transformation in control theory Baranyi and Yam proposed the TP model transformation as a new concept in quasi-LPV (qLPV) based control, which plays a central role in the highly desirable bridging between identification and polytopic systems theories. It is also used as a TS (Tak ...
Bounded operators on Hilbert spaces
The factorization can be extended to a
bounded operator
In functional analysis and operator theory, a bounded linear operator is a linear transformation L : X \to Y between topological vector spaces (TVSs) X and Y that maps bounded subsets of X to bounded subsets of Y.
If X and Y are normed vector ...
''M'' on a separable Hilbert space ''H''. Namely, for any bounded operator ''M'', there exist a
partial isometry In functional analysis a partial isometry is a linear map between Hilbert spaces such that it is an isometry on the orthogonal complement of its kernel.
The orthogonal complement of its kernel is called the initial subspace and its range is called ...
''U'', a unitary ''V'', a measure space (''X'', ''μ''), and a non-negative measurable ''f'' such that
:
where
is the
multiplication by ''f'' on ''L''
2(''X'', ''μ'').
This can be shown by mimicking the linear algebraic argument for the matricial case above. ''VT''
''f''''V''* is the unique positive square root of ''M*M'', as given by the
Borel functional calculus
In functional analysis, a branch of mathematics, the Borel functional calculus is a '' functional calculus'' (that is, an assignment of operators from commutative algebras to functions defined on their spectra), which has particularly broad scop ...
for
self-adjoint operator
In mathematics, a self-adjoint operator on an infinite-dimensional complex vector space ''V'' with inner product \langle\cdot,\cdot\rangle (equivalently, a Hermitian operator in the finite-dimensional case) is a linear map ''A'' (from ''V'' to itse ...
s. The reason why ''U'' need not be unitary is because, unlike the finite-dimensional case, given an isometry ''U''
1 with nontrivial kernel, a suitable ''U''
2 may not be found such that
:
is a unitary operator.
As for matrices, the singular value factorization is equivalent to the
polar decomposition
In mathematics, the polar decomposition of a square real or complex matrix A is a factorization of the form A = U P, where U is an orthogonal matrix and P is a positive semi-definite symmetric matrix (U is a unitary matrix and P is a positive se ...
for operators: we can simply write
:
and notice that ''U V*'' is still a partial isometry while ''VT''
''f''''V''* is positive.
Singular values and compact operators
The notion of singular values and left/right-singular vectors can be extended to
compact operator on Hilbert space
In the mathematical discipline of functional analysis, the concept of a compact operator on Hilbert space is an extension of the concept of a matrix acting on a finite-dimensional vector space; in Hilbert space, compact operators are precisely the ...
as they have a discrete spectrum. If is compact, every non-zero in its spectrum is an eigenvalue. Furthermore, a compact self-adjoint operator can be diagonalized by its eigenvectors. If is compact, so is . Applying the diagonalization result, the unitary image of its positive square root has a set of orthonormal eigenvectors corresponding to strictly positive eigenvalues . For any ,
:
where the series converges in the norm topology on . Notice how this resembles the expression from the finite-dimensional case. are called the singular values of . (resp. ) can be considered the left-singular (resp. right-singular) vectors of .
Compact operators on a Hilbert space are the closure of
finite-rank operator
In functional analysis, a branch of mathematics, a finite-rank operator is a bounded linear operator between Banach spaces whose range is finite-dimensional.
Finite-rank operators on a Hilbert space A canonical form
Finite-rank operators are ...
s in the uniform operator topology. The above series expression gives an explicit such representation. An immediate consequence of this is:
:Theorem. is compact if and only if is compact.
History
The singular value decomposition was originally developed by
differential geometers
Differential geometry is a mathematical discipline that studies the geometry of smooth shapes and smooth spaces, otherwise known as smooth manifolds. It uses the techniques of differential calculus, integral calculus, linear algebra and multil ...
, who wished to determine whether a real
bilinear form
In mathematics, a bilinear form is a bilinear map on a vector space (the elements of which are called '' vectors'') over a field ''K'' (the elements of which are called ''scalars''). In other words, a bilinear form is a function that is linear i ...
could be made equal to another by independent orthogonal transformations of the two spaces it acts on.
Eugenio Beltrami and
Camille Jordan
Marie Ennemond Camille Jordan (; 5 January 1838 – 22 January 1922) was a French mathematician, known both for his foundational work in group theory and for his influential ''Cours d'analyse''.
Biography
Jordan was born in Lyon and educated at ...
discovered independently, in 1873 and 1874 respectively, that the singular values of the bilinear forms, represented as a matrix, form a
complete set of
invariant
Invariant and invariance may refer to:
Computer science
* Invariant (computer science), an expression whose value doesn't change during program execution
** Loop invariant, a property of a program loop that is true before (and after) each iteratio ...
s for bilinear forms under orthogonal substitutions.
James Joseph Sylvester
James Joseph Sylvester (3 September 1814 – 15 March 1897) was an English mathematician. He made fundamental contributions to matrix theory, invariant theory, number theory, partition theory, and combinatorics. He played a leadership ro ...
also arrived at the singular value decomposition for real square matrices in 1889, apparently independently of both Beltrami and Jordan. Sylvester called the singular values the ''canonical multipliers'' of the matrix ''A''. The fourth mathematician to discover the singular value decomposition independently is
Autonne in 1915, who arrived at it via the
polar decomposition
In mathematics, the polar decomposition of a square real or complex matrix A is a factorization of the form A = U P, where U is an orthogonal matrix and P is a positive semi-definite symmetric matrix (U is a unitary matrix and P is a positive se ...
. The first proof of the singular value decomposition for rectangular and complex matrices seems to be by
Carl Eckart
Carl Henry Eckart (May 4, 1902 – October 23, 1973) was an American physicist, physical oceanographer, geophysicist, and administrator. He co-developed the Wigner–Eckart theorem and is also known for the Eckart conditions in quantum mechanics ...
and
Gale J. Young in 1936; they saw it as a generalization of the
principal axis transformation for
Hermitian matrices
In mathematics, a Hermitian matrix (or self-adjoint matrix) is a complex square matrix that is equal to its own conjugate transpose—that is, the element in the -th row and -th column is equal to the complex conjugate of the element in the -th ...
.
In 1907,
Erhard Schmidt
Erhard Schmidt (13 January 1876 – 6 December 1959) was a Baltic German mathematician whose work significantly influenced the direction of mathematics in the twentieth century. Schmidt was born in Tartu (german: link=no, Dorpat), in the Govern ...
defined an analog of singular values for
integral operator
An integral operator is an operator that involves integration. Special instances are:
* The operator of integration itself, denoted by the integral symbol
* Integral linear operators, which are linear operators induced by bilinear forms invol ...
s (which are compact, under some weak technical assumptions); it seems he was unaware of the parallel work on singular values of finite matrices. This theory was further developed by
Émile Picard
Charles Émile Picard (; 24 July 1856 – 11 December 1941) was a French mathematician. He was elected the fifteenth member to occupy seat 1 of the Académie française in 1924.
Life
He was born in Paris on 24 July 1856 and educated there at t ...
in 1910, who is the first to call the numbers
''singular values'' (or in French, ''valeurs singulières'').
Practical methods for computing the SVD date back to
Kogbetliantz in 1954–1955 and
Hestenes in 1958, resembling closely the
Jacobi eigenvalue algorithm In numerical linear algebra, the Jacobi eigenvalue algorithm is an iterative method for the calculation of the eigenvalues and eigenvectors of a real symmetric matrix (a process known as diagonalization). It is named after Carl Gustav Jacob Jacobi, ...
, which uses plane rotations or
Givens rotation In numerical linear algebra, a Givens rotation is a rotation in the plane spanned by two coordinates axes. Givens rotations are named after Wallace Givens, who introduced them to numerical analysts in the 1950s while he was working at Argonne Nation ...
s. However, these were replaced by the method of
Gene Golub
Gene Howard Golub (February 29, 1932 – November 16, 2007), was an American numerical analyst who taught at Stanford University as Fletcher Jones Professor of Computer Science and held a courtesy appointment in electrical engineering.
Person ...
and
William Kahan
William "Velvel" Morton Kahan (born June 5, 1933) is a Canadian mathematician and computer scientist, who received the Turing Award in 1989 for "''his fundamental contributions to numerical analysis''",
was named an ACM Fellow in 1994, and inducte ...
published in 1965, which uses
Householder transformation
In linear algebra, a Householder transformation (also known as a Householder reflection or elementary reflector) is a linear transformation that describes a reflection about a plane or hyperplane containing the origin. The Householder transformati ...
s or reflections. In 1970, Golub and Christian Reinsch
published a variant of the Golub/Kahan algorithm that is still the one most-used today.
See also
Notes
References
*
*
*
*
*
*
*
* Halldor, Bjornsson and Venegas, Silvia A. (1997)
"A manual for EOF and SVD analyses of climate data" McGill University, CCGCR Report No. 97-1, Montréal, Québec, 52pp.
*
*
*
*
*
*
*
*
External links
{{DEFAULTSORT:Singular Value Decomposition
Linear algebra
Numerical linear algebra
Matrix theory
Matrix decompositions
Functional analysis