The R Project SVN R

% File src/library/base/man/svd.Rd
% Part of the R package, https://www.R-project.org
% Copyright 1995-2020 R Core Team
% Distributed under GPL 2 or later

\name{svd}
\alias{svd}
\alias{La.svd}
\title{Singular Value Decomposition of a Matrix}
\description{
  Compute the singular-value decomposition of a rectangular matrix.
}
\usage{
svd(x, nu = min(n, p), nv = min(n, p), LINPACK = FALSE)

La.svd(x, nu = min(n, p), nv = min(n, p))
}
\arguments{
  \item{x}{a numeric or complex matrix whose SVD decomposition
    is to be computed.  Logical matrices are coerced to numeric.}
  \item{nu}{the number of left  singular vectors to be computed.
    This must between \code{0} and \code{n = nrow(x)}.}
  \item{nv}{the number of right singular vectors to be computed.
    This must be between \code{0} and \code{p = ncol(x)}.}
  \item{LINPACK}{logical.  Defunct and ignored with a warning.}
}
\details{
  The singular value decomposition plays an important role in many
  statistical techniques.  \code{svd} and \code{La.svd} provide two
  interfaces which differ in their return values.

  Computing the singular vectors is the slow part for large matrices.
  The computation will be more efficient if both \code{nu <= min(n, p)}
  and \code{nv <= min(n, p)}, and even more so if both are zero.

  Unsuccessful results from the underlying LAPACK code will result in an
  error giving a positive error code (most often \code{1}): these can
  only be interpreted by detailed study of the FORTRAN code but mean
  that the algorithm failed to converge.
}
\value{
  The SVD decomposition of the matrix as computed by LAPACK, \deqn{
  \bold{X = U D V'},} where \eqn{\bold{U}} and \eqn{\bold{V}} are
  orthogonal, \eqn{\bold{V'}} means \emph{V transposed} (and conjugated
  for complex input), and \eqn{\bold{D}} is a diagonal matrix with the
  (non-negative) singular values \eqn{D_{ii}}{D[i,i]} in decreasing
  order.  Equivalently, \eqn{\bold{D = U' X V}}, which is verified in
  the examples.

  The returned value is a list with components
  \item{d}{a vector containing the singular values of \code{x}, of
    length \code{min(n, p)}, sorted decreasingly.}
  \item{u}{a matrix whose columns contain the left singular vectors of
    \code{x}, present if \code{nu > 0}.  Dimension \code{c(n, nu)}.}
  \item{v}{a matrix whose columns contain the right singular vectors of
    \code{x}, present if \code{nv > 0}.  Dimension \code{c(p, nv)}.}

  Recall that the singular vectors are only defined up to sign (a
  constant of modulus one in the complex case).  If a left singular
  vector has its sign changed, changing the sign of the corresponding
  right vector gives an equivalent decomposition.

  For \code{La.svd} the return value replaces \code{v} by \code{vt}, the
  (conjugated if complex) transpose of \code{v}.
}
\source{
  The main functions used are the LAPACK routines \code{DGESDD} and
  \code{ZGESDD}.

  LAPACK is from \url{https://www.netlib.org/lapack/} and its guide is
  listed in the references.
}
\references{
  Anderson. E. and ten others (1999)
  \emph{LAPACK Users' Guide}. Third Edition. SIAM.\cr
  Available on-line at
  \url{https://www.netlib.org/lapack/lug/lapack_lug.html}.

  The \href{https://en.wikipedia.org/wiki/Singular-value_decomposition}{%
    \sQuote{Singular-value decomposition}} Wikipedia article.

  Becker, R. A., Chambers, J. M. and Wilks, A. R. (1988)
  \emph{The New S Language}.
  Wadsworth & Brooks/Cole.
}

\seealso{
  \code{\link{eigen}}, \code{\link{qr}}.
}
\examples{
hilbert <- function(n) { i <- 1:n; 1 / outer(i - 1, i, "+") }
X <- hilbert(9)[, 1:6]
(s <- svd(X))
D <- diag(s$d)
s$u \%*\% D \%*\% t(s$v) #  X = U D V'
t(s$u) \%*\% X \%*\% s$v #  D = U' X V
}
\keyword{algebra}
\keyword{array}