singular value

TheInfoList

In
mathematics Mathematics (from Greek: ) includes the study of such topics as numbers ( and ), formulas and related structures (), shapes and spaces in which they are contained (), and quantities and their changes ( and ). There is no general consensus abo ...
, in particular
functional analysis 200px, One of the possible modes of vibration of an idealized circular drum head. These modes are eigenfunctions of a linear operator on a function space, a common construction in functional analysis. Functional analysis is a branch of mathemat ...
, the singular values, or ''s''-numbers of a
compact operator In functional analysis Image:Drum vibration mode12.gif, 200px, One of the possible modes of vibration of an idealized circular drum head. These modes are eigenfunctions of a linear operator on a function space, a common construction in functional a ...
acting between
Hilbert space In mathematics Mathematics (from Greek: ) includes the study of such topics as numbers ( and ), formulas and related structures (), shapes and spaces in which they are contained (), and quantities and their changes ( and ). There is no ge ...
s ''X'' and ''Y'', are the square roots of non-negative
eigenvalue 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 an ...

s of the self-adjoint operator (where ''T'' denotes the adjoint of ''T''). The singular values are non-negative
real number In mathematics Mathematics (from Ancient Greek, Greek: ) includes the study of such topics as quantity (number theory), mathematical structure, structure (algebra), space (geometry), and calculus, change (mathematical analysis, analysis). ...
s, usually listed in decreasing order (''s''1(''T''), ''s''2(''T''), …). The largest singular value ''s''1(''T'') is equal to the
operator norm In mathematics Mathematics (from Ancient Greek, Greek: ) includes the study of such topics as quantity (number theory), mathematical structure, structure (algebra), space (geometry), and calculus, change (mathematical analysis, analysis). It h ...
of ''T'' (see
Min-max theorem In linear algebra and functional analysis, the min-max theorem, or variational theorem, or Courant–Fischer–Weyl min-max principle, is a result that gives a variational characterization of Eigenvalues and eigenvectors, eigenvalues of ...
). If ''T'' acts on euclidean space R''n'', there is a simple geometric interpretation for the singular values: Consider the image by ''T'' of the N-sphere, unit sphere; this is an ellipsoid, and the lengths of its semi-axes are the singular values of ''T'' (the figure provides an example in R''2''). The singular values are the absolute values of the eigenvalues of a normal matrix ''A'', because the spectral theorem can be applied to obtain unitary diagonalization of ''A'' as . Therefore, Most normed linear space, norms on Hilbert space operators studied are defined using ''s''-numbers. For example, the Ky Fan-''k''-norm is the sum of first ''k'' singular values, the trace norm is the sum of all singular values, and the Schatten norm is the ''p''th root of the sum of the ''p''th powers of the singular values. Note that each norm is defined only on a special class of operators, hence ''s''-numbers are useful in classifying different operators. In the finite-dimensional case, a matrix (mathematics), matrix can always be decomposed in the form ''U''Σ''V'', where ''U'' and ''V'' are unitary matrix, unitary matrices and Σ is a diagonal matrix with the singular values lying on the diagonal. This is the singular value decomposition.

# Basic properties

For $A \in \mathbb^$, and $i = 1,2, \ldots, \min \$. Min-max theorem#Min-max principle for singular values, Min-max theorem for singular values. Here $U: \dim\left(U\right) = i$ is a subspace of $\mathbb^n$ of dimension $i$. :$\begin \sigma_i\left(A\right) &= \min_ \max_ \left\, Ax \right\, _2. \\ \sigma_i\left(A\right) &= \max_ \min_ \left\, Ax \right\, _2. \end$ Matrix transpose and conjugate do not alter singular values. :$\sigma_i\left(A\right) = \sigma_i\left\left(A^\textsf\right\right) = \sigma_i\left\left(A^*\right\right) = \sigma_i\left\left(\bar\right\right).$ For any unitary $U \in \mathbb^, V \in \mathbb^.$ :$\sigma_i\left(A\right) = \sigma_i\left(UAV\right).$ Relation to eigenvalues: :$\sigma_i^2\left(A\right) = \lambda_i\left\left(AA^*\right\right) = \lambda_i\left\left(A^*A\right\right).$

## Singular values of sub-matrices

For $A \in \mathbb^.$ # Let $B$ denote $A$ with one of its rows ''or'' columns deleted. Then #: $\sigma_\left(A\right) \leq \sigma_i \left(B\right) \leq \sigma_i\left(A\right)$ # Let $B$ denote $A$ with one of its rows ''and'' columns deleted. Then #: $\sigma_\left(A\right) \leq \sigma_i \left(B\right) \leq \sigma_i\left(A\right)$ # Let $B$ denote an $\left(m-k\right)\times\left(n-l\right)$ submatrix of $A$. Then #: $\sigma_\left(A\right) \leq \sigma_i \left(B\right) \leq \sigma_i\left(A\right)$

## Singular values of ''A'' + ''B''

For $A, B \in \mathbb^$ # $\sum_^ \sigma_i\left(A + B\right) \leq \sum_^ \sigma_i\left(A\right) + \sigma_i\left(B\right), \quad k=\min \$ # $\sigma_\left(A + B\right) \leq \sigma_i\left(A\right) + \sigma_j\left(B\right). \quad i,j\in\mathbb,\ i + j - 1 \leq \min \$

## Singular values of ''AB''

For $A, B \in \mathbb^$ # $\begin \prod_^ \sigma_i\left(A\right) \sigma_i\left(B\right) &\leq \prod_^ \sigma_i\left(AB\right) \\ \prod_^k \sigma_i\left(AB\right) &\leq \prod_^k \sigma_i\left(A\right) \sigma_i\left(B\right), \\ \sum_^k \sigma_i^p\left(AB\right) &\leq \sum_^k \sigma_i^p\left(A\right) \sigma_i^p\left(B\right), \end$ # $\sigma_n\left(A\right) \sigma_i\left(B\right) \leq \sigma_i \left(AB\right) \leq \sigma_1\left(A\right) \sigma_i\left(B\right) \quad i = 1, 2, \ldots, n.$ For $A, B \in \mathbb^$ :$2 \sigma_i\left(A B^*\right) \leq \sigma_i \left\left(A^* A + B^* B\right\right), \quad i = 1, 2, \ldots, n.$

## Singular values and eigenvalues

For $A \in \mathbb^$. # See #: $\lambda_i\left\left(A + A^*\right\right) \leq 2 \sigma_i\left(A\right), \quad i = 1, 2, \ldots, n.$ # Assume $\left, \lambda_1\left(A\right)\ \geq \cdots \geq \left, \lambda_n\left(A\right)\$. Then for $k = 1, 2, \ldots, n$: ## Weyl's inequality#Weyl's inequality in matrix theory, Weyl's theorem ##: $\prod_^k \left, \lambda_i\left(A\right)\ \leq \prod_^ \sigma_i\left(A\right).$ ## For $p>0$. ##: $\sum_^k \left, \lambda_i^p\left(A\right)\ \leq \sum_^ \sigma_i^p\left(A\right).$

# History

This concept was introduced by Erhard Schmidt in 1907. Schmidt called singular values "eigenvalues" at that time. The name "singular value" was first quoted by Smithies in 1937. In 1957, Allahverdiev proved the following characterization of the ''n''th ''s''-number:Israel Gohberg, I. C. Gohberg and Mark Krein, M. G. Krein. Introduction to the Theory of Linear Non-selfadjoint Operators. American Mathematical Society, Providence, R.I.,1969. Translated from the Russian by A. Feinstein. Translations of Mathematical Monographs, Vol. 18. : $s_n\left(T\right) = \inf\big\.$ This formulation made it possible to extend the notion of ''s''-numbers to operators in Banach space.