NAG Library Chapter Introduction
F04 – Simultaneous Linear Equations
1 Scope of the Chapter
This chapter is concerned with the solution of the matrix equation , where may be a single vector or a matrix of multiple right-hand sides. The matrix may be real, complex, symmetric, Hermitian, positive definite, positive definite Toeplitz or banded. It may also be rectangular, in which case a least squares solution is obtained.
Much of the functionality of this chapter has been superseded by routines from Chapters F07
(LAPACK routines) as those chapters have grown and have included driver and expert driver routines.
For a general introduction to sparse systems of equations, see the F11 Chapter Introduction
, which provides routines for large sparse systems.
Some routines for sparse problems are also included in this chapter; they are described in Section 3.4
2 Background to the Problems
A set of linear equations may be written in the form
where the known matrix
, with real or complex coefficients, is of size
columns), the known right-hand vector
rows and one column), and the required solution vector
rows and one column). There may also be
, on the right-hand side and the equations may then be written as
the required matrix
having as its
columns the solutions of
. Some routines deal with the latter case, but for clarity only the case
is discussed here.
The most common problem, the determination of the unique solution of
, occurs when
is not singular, that is
. This is discussed in Section 2.1
below. The next most common problem, discussed in Section 2.2
below, is the determination of the least squares solution of
, i.e., the determination of the vector
which minimizes the norm of the residual vector
. All other cases are rank deficient, and they are treated in Section 2.3
2.1 Unique Solution of
Most routines in this chapter solve this particular problem. The computation starts with the triangular decomposition
are respectively lower and upper triangular matrices and
is a permutation matrix, chosen so as to ensure that the decomposition is numerically stable. The solution is then obtained by solving in succession the simpler equations
the first by forward-substitution and the second by back-substitution.
If is real symmetric and positive definite, , while if is complex Hermitian and positive definite, ; in both these cases is the identity matrix (i.e., no permutations are necessary). In all other cases either or has unit diagonal elements.
Due to rounding errors the computed ‘solution’
, say, is only an approximation to the true solution . This approximation will sometimes be satisfactory, agreeing with to several figures, but if the problem is ill-conditioned then and may have few or even no figures in common, and at this stage there is no means of estimating the ‘accuracy’ of .
There are three possible approaches to estimating the accuracy of a computed solution.
One way to do so, and to ‘correct’
when this is meaningful (see next paragraph), involves computing the residual vector in extended precision arithmetic, and obtaining a correction vector by solving . The new approximate solution is usually more accurate and the correcting process is repeated until (a) further corrections are negligible or (b) they show no further decrease.
It must be emphasized that the ‘true’ solution may not be meaningful, that is correct to all figures quoted, if the elements of and are known with certainty only to say figures, where is less than full precision.
The first correction vector will then give some useful information about the number of figures in the ‘solution’ which probably remain unchanged with respect to maximum possible uncertainties in the coefficients.
approach to assessing the accuracy of the solution is to compute or estimate the condition number
, defined as
Roughly speaking, errors or uncertainties in
may be amplified in the solution by a factor
. Thus, for example, if the data in
are only accurate to
, then the solution cannot be guaranteed to have more than
correct digits. If
, the solution may have no meaningful digits.
To be more precise, suppose that
represent perturbations to the matrices
which cause a perturbation
in the solution. We can define measures of the relative sizes of the perturbations in
and then the bound effectively simplifies to
Hence, if we know
, we can compute a bound on the relative errors in the solution. Note that
are defined in terms of the norms of
contains elements of widely differing magnitude, then
will be dominated by the errors in the larger elements, and
will give no information about the relative accuracy of smaller elements of
way to obtain useful information about the accuracy of a solution is to solve two sets of equations, one with the given coefficients, which are assumed to be known with certainty to figures, and one with the coefficients rounded to () figures, and to count the number of figures to which the two solutions agree. In ill-conditioned problems this can be surprisingly small and even zero.
2.2 The Least Squares Solution of , ,
The least squares solution is the vector
which minimizes the sum of the squares of the residuals,
The solution is obtained in two steps.
||Householder transformations are used to reduce to ‘simpler form’ via the equation , where has the appearance
with a nonsingular upper triangular by matrix and a zero matrix of shape by . Similar operations convert to , where
with having rows and having () rows.
||The required least squares solution is obtained by back-substitution in the equation
Again due to rounding errors the computed
is only an approximation to the required
, but as in Section 2.1
, this can be improved by ‘iterative refinement’. The first correction
is the solution of the least squares problem
and since the matrix
is unchanged, this computation takes less time than that of the original
. The process can be repeated until further corrections are (a)
negligible or (b)
show no further decrease.
2.3 Rank-deficient Cases
If, in the least squares problem just discussed, , then a least squares solution exists but it is not unique. In this situation it is usual to ask for the least squares solution ‘of minimal length’, i.e., the vector which minimizes , among all those for which is a minimum.
This can be computed from the Singular Value Decomposition (SVD) of
, in which
is factorized as
matrix with orthonormal columns,
orthogonal matrix and
diagonal matrix. The diagonal elements of
are called the ‘singular values’ of
; they are non-negative and can be arranged in decreasing order of magnitude:
The columns of
are called respectively the left and right singular vectors of
. If the singular values
are zero or negligible, but
is not negligible, then the rank of
is taken to be
(see also Section 2.4
) and the minimal length least squares solution of
is given by
is the diagonal matrix with diagonal elements
The SVD may also be used to find solutions to the homogeneous system of equations
. Such solutions exist if and only if
, and are given by
are arbitrary numbers and the
are the columns of
which correspond to negligible elements of
The general solution to the rank-deficient least squares problem is given by , where is the minimal length least squares solution and is any solution of the homogeneous system of equations .
2.4 The Rank of a Matrix
In theory the rank is if elements of the diagonal matrix of the singular value decomposition are exactly zero. In practice, due to rounding and/or experimental errors, some of these elements have very small values which usually can and should be treated as zero.
For example, the following
matrix has rank
in exact arithmetic:
On a computer with
decimal digits of precision the computed singular values were
and the rank would be correctly taken to be
It is not, however, always certain that small computed singular values are really zero. With the
Hilbert matrix, for example, where
, the singular values are
Here there is no clear cut-off between small (i.e., negligible) singular values and larger ones. In fact, in exact arithmetic, the matrix is known to have full rank and none of its singular values is zero. On a computer with
decimal digits of precision, the matrix is effectively singular, but should its rank be taken to be
It is therefore impossible to give an infallible rule, but generally the rank can be taken to be the number of singular values which are neither zero nor very small compared with other singular values. For example, if there is a sharp decrease in singular values from numbers of order unity to numbers of order , then the latter will almost certainly be zero in a machine in which significant decimal figures is the maximum accuracy. Similarly for a least squares problem in which the data is known to about four significant figures and the largest singular value is of order unity then a singular value of order or less should almost certainly be regarded as zero.
It should be emphasized that rank determination and least squares solutions can be sensitive to the scaling of the matrix. If at all possible the units of measurement should be chosen so that the elements of the matrix have data errors of approximately equal magnitude.
2.5 Generalized Linear Least Squares Problems
The simple type of linear least squares problem described in Section 2.2
can be generalized in various ways.
||Linear least squares problems with equality constraints:
where is by and is by , with . The equations may be regarded as a set of equality constraints on the problem of minimizing . Alternatively the problem may be regarded as solving an overdetermined system of equations
where some of the equations (those involving ) are to be solved exactly, and the others (those involving ) are to be solved in a least squares sense. The problem has a unique solution on the assumptions that has full row rank and the matrix has full column rank . (For linear least squares problems with inequality constraints, refer to Chapter E04.)
||General Gauss–Markov linear model problems:
where is by and is by , with . When , the problem reduces to an ordinary linear least squares problem. When is square and nonsingular, it is equivalent to a weighted linear least squares problem:
The problem has a unique solution on the assumptions that has full column rank , and the matrix has full row rank .
2.6 Calculating the Inverse of a Matrix
The routines in this chapter can also be used to calculate the inverse of a square matrix
by solving the equation
is the identity matrix. However, solving the equations
by calculation of the inverse of the coefficient matrix
, i.e., by
, is definitely not recommended
Similar remarks apply to the calculation of the pseudo-inverse of a singular or rectangular matrix.
2.7 Estimating the 1-norm of a Matrix
The 1-norm of a matrix
is defined to be:
Typically it is useful to calculate the condition number of a matrix with respect to the solution of linear equations, or inversion. The higher the condition number the less accuracy might be expected from a numerical computation. A condition number for the solution of linear equations is . Since this might be a relatively expensive computation it often suffices to estimate the norm of each matrix.
3 Recommendations on Choice and Use of Available Routines
See also Section 3
in the F07 Chapter Introduction for recommendations on the choice of available routines from that chapter.
3.1 Black Box and General Purpose Routines
Most of the routines in this chapter are categorised either as Black Box routines or general purpose routines.
Black Box routines solve the equations , for , in a single call with the matrix and the right-hand sides, , being supplied as data. These are the simplest routines to use and are suitable when all the right-hand sides are known in advance and do not occupy too much storage.
General purpose routines, in general, require a previous call to a routine in Chapters F01
to factorize the matrix
. This factorization can then be used repeatedly to solve the equations for one or more right-hand sides which may be generated in the course of the computation. The Black Box routines simply call a factorization routine and then a general purpose routine to solve the equations.
The routine F04QAF
which uses an iterative method for sparse systems of equations does not fit easily into this categorization, but is classified as a general purpose routine in the decision trees and indexes.
3.2 Systems of Linear Equations
Most of the routines in this chapter solve linear equations
and a unique solution is expected (see Section 2.1
). The matrix
may be ‘general’ real or complex, or may have special structure or properties, e.g., it may be banded, tridiagonal, almost block-diagonal, sparse, symmetric, Hermitian, positive definite (or various combinations of these).
It must be emphasized that it is a waste of computer time and space to use an inappropriate routine, for example one for the complex case when the equations are real. It is also unsatisfactory to use the special routines for a positive definite matrix if this property is not known in advance.
Routines are given for calculating the approximate solution
, that is solving the linear equations just once, and also for obtaining the accurate solution
by successive iterative corrections of this first approximation using additional precision arithmetic, as described in Section 2.1
. The latter, of course, are more costly in terms of time and storage, since each correction involves the solution of
sets of linear equations and since the original
decomposition must be stored together with the first and successively corrected approximations to the solution. In practice the storage requirements for the ‘corrected’ routines are about double those of the ‘approximate’ routines, though the extra computer time may not be prohibitive since the same matrix and the same
decomposition is used in every linear equation solution.
A number of the Black Box routines in this chapter return estimates of the condition number and the forward error, along with the solution of the equations. But for those routines that do not return a condition estimate two routines are provided – F04YDF
for real matrices, F04ZDF
for complex matrices – which can return a cheap but reliable estimate of
, and hence an estimate of the condition number
(see Section 2.1
). These routines can also be used in conjunction with most of the linear equation solving routines in Chapter F11
: further advice is given in the routine documents.
Other routines for solving linear equation systems, computing inverse matrices, and estimating condition numbers can be found in Chapter F07
, which contains LAPACK software.
3.3 Linear Least Squares Problems
The majority of the routines for solving linear least squares problems are to be found in Chapter F08
For the case described in Section 2.2
and a unique least squares solution is expected, there are two routines for a general real
, one of which (F04JGF
) computes a first approximation and the other (F04AMF
) computes iterative corrections. If it transpires that
, so that the least squares solution is not unique, then F04AMF
takes a failure exit, but F04JGF
proceeds to compute the minimal length
solution by using the SVD (see below).
If is expected to be of less than full rank then one of the routines for calculating the minimal length solution may be used.
For the use of the SVD is not significantly more expensive than the use of routines based upon the factorization.
Problems with linear equality constraints
can be solved by
(for real data) or by F08ZNF (ZGGLSE)
(for complex data),
provided that the problems are of full rank. Problems with linear inequality constraints
can be solved by E04NCF/E04NCA
in Chapter E04
General Gauss–Markov linear model problems, as formulated in Section 2.5
, can be solved by
(for real data) or by F08ZPF (ZGGGLM)
(for complex data).
3.4 Sparse Matrix Routines
Routines specifically for sparse matrices are appropriate only when the number of nonzero elements is very small, less than, say, 10% of the elements of , and the matrix does not have a relatively small band width.
contains routines for both the direct and iterative solution of sparse linear systems. There are two routines in Chapter F04
for solving sparse linear equations (F04AXF
utilizes a factorization of the matrix
obtained from F01BRF
, while F04QAF
uses an iterative technique and requires a user-supplied function to compute matrix-vector products
for any given vector
solves sparse least squares problems by an iterative technique, and also allows the solution of damped (regularized) least squares problems (see the routine document for details).
4 Decision Trees
The name of the routine (if any) that should be used to factorize the matrix is given in brackets after the name of the routine for solving the equations.
Tree 1: Black Box routines for unique solution of (Real matrix)
Tree 2: Black Box routines for unique solution of (Complex matrix)
Tree 3: General purpose routines for unique solution of (Real matrix)
Tree 4: General purpose routines for unique solution of (Complex matrix)
Tree 5: General purpose routines for least squares and homogeneous equations (without constraints)
there are also routines in Chapter F08
for solving least squares problems.
Note 1: also returns an estimate of the condition number and the forward error.
Note 2: also returns an estimate of the condition number, the forward error and the backward error. Requires additional workspace.
Note 3: for a single right-hand side only.
5 Functionality Index
|Black Box routines, Ax = b,|| |
| complex general band matrix|| ||F04CBF|
| complex general tridiagonal matrix|| ||F04CCF|
| complex Hermitian matrix,|| |
| complex Hermitian positive definite band matrix|| ||F04CFF|
| complex Hermitian positive definite matrix,|| |
| complex Hermitian positive definite tridiagonal matrix|| ||F04CGF|
| complex symmetric matrix,|| |
| real general band matrix|| ||F04BBF|
| multiple right-hand sides,|| |
| iterative refinement using additional precision|| ||F04AEF|
| multiple right-hand sides, standard precision|| ||F04BAF|
| iterative refinement using additional precision|| ||F04ATF|
| real general tridiagonal matrix|| ||F04BCF|
| real symmetric positive definite band matrix|| ||F04BFF|
| real symmetric positive definite matrix,|| |
| multiple right-hand sides,|| |
| iterative refinement using additional precision|| ||F04ABF|
| multiple right-hand sides, standard precision|| ||F04BDF|
| iterative refinement using additional precision|| ||F04ASF|
| real symmetric positive definite Toeplitz matrix,|| |
| general right-hand side|| ||F04FFF|
| real symmetric positive definite tridiagonal matrix|| ||F04BGF|
|General Purpose routines, Ax = b,|| |
| real almost block-diagonal matrix|| ||F04LHF|
| real band symmetric positive definite matrix, variable bandwidth|| ||F04MCF|
| real symmetric positive definite Toeplitz matrix,|| |
| general right-hand side, update solution|| ||F04MFF|
| Yule–Walker equations, update solution|| ||F04MEF|
| real tridiagonal matrix|| ||F04LEF|
|Least squares and Homogeneous Equations,|| |
| m ≥ n, rank = n or minimal solution|| ||F04JGF|
| rank = n, iterative refinement|| ||F04AMF|
| complex rectangular matrix,|| |
| norm and condition number estimation|| ||F04ZDF|
| covariance matrix for linear least squares problems|| ||F04YAF|
| norm and condition number estimation|| ||F04YDF|
6 Auxiliary Routines Associated with Library Routine Parameters
7 Routines Withdrawn or Scheduled for Withdrawal
The following lists all those routines that have been withdrawn since Mark 18 of the Library or are scheduled for withdrawal at one of the next two marks.
Golub G H and Van Loan C F (1996) Matrix Computations (3rd Edition) Johns Hopkins University Press, Baltimore
Lawson C L and Hanson R J (1974) Solving Least Squares Problems Prentice–Hall
Wilkinson J H and Reinsch C (1971) Handbook for Automatic Computation II, Linear Algebra Springer–Verlag