This section describes functions from file la.ct.
[y] = axpy(a,x)
[y] = axpy(a,x) is the equivalent of y = y + a*x
for scalar a and (real or complex) arrays x and y.
axpy calls a fast BLAS-1 routine if available.
Error codes:
-1: Input/output arg y is not real/complex array
-2: First input arg a is not a scalar
-3: Second input arg x is not real/complex array
-4: Second input arg x has size incompatible with y
-5: If y is real a and x must also be real
-6: If y is complex x must also be complex
[B] = chol(A)
B=chol(A) returns the Cholesky decomposition of a
positive definite square matrix A: B'**B == A.
A may be integer, real or complex valued.
Only the upper half of A is referenced, the lower half
is assumed to be Hermitian symmetric.
See also:
linsolve,
LU,
matprod,
det,
eig,
inv.
Error codes:
-1: Input arg is not a matrix
-2: Input matrix is not square
-3: Input matrix is not positive definite
[d] = det(A)
det(A) returns the determinant of a square matrix A.
A may be integer, real or complex valued.
If A is scalar, it is returned as such.
Error codes:
-1: Nonnumeric input arg
-2: Input array is not a matrix
-3: Input matrix is not square
-4: Singular matrix
[D;V] = eig(A)
eig(A) returns the eigenvalues of a square matrix A.
[D,V] = eig(A) returns the eigenvalues in D and the
right eigenvectors as columns of V. The eigenvectors
satisfy A**V == D*V.
See also:
inv,
LU.
Error codes:
1: Failed to converge
-1: Input arg is not an array
-2: Input array is not a matrix
-3: Input matrix is not square
-4: Internal error
[A] = eye(n)
eye(n) returns the (integer) unit matrix of order n.
n must be a non-negative scalar integer.
eye(V) where V is a two-element integer vector with
both elements equal and positive works also, thus
you can also use eye(size(A)).
See also:
ones,
inv.
Error codes:
-1: Argument not an integer or IntArray
-2: Negative dimension
-3: IntArray rank not 1
-4: IntArray length not 2
-5: Integer vector elements are unequal
[B] = inv(A)
inv(A) returns the inverse of a square matrix A.
A may be integer, real or complex valued.
A may also be a scalar, in which case its reciprocal
is returned.
See also:
linsolve,
LU,
chol,
matprod,
det,
eig.
Error codes:
-1: Nonnumeric input arg
-2: Input array is not a matrix
-3: Input matrix is not square
-4: Singular matrix
-5: Singular matrix
[x] = linsolve(A,b)
linsolve(A,b) solves the linear system A**x == b.
If A is square, the result x is roughly the same as
computing inv(A)**b (however, using linsolve is
faster and numerically more accurate). If A is not
square, a least-square problem is solved. If the system
is overdetermined, the solution x minimizes the quantity
|A**x - b|. If the system is underdetermined, the
solution x minimizes |x| among all x that satisfy
A**x==b.
The second argument may be a vector or a matrix.
If it is a matrix, several linear systems are effectively
solved simultaneously.
See also:
inv,
LU,
eig,
SVD.
Error codes:
-1: First input arg is not an array
-2: First input arg is not a matrix
-3: Second input arg is not an array
-4: Second input arg is not a vector or matrix
-5: Incompatible dimensions in first/second args
-6: Matrix must be square
-7: Singular matrix
-8: Internal error
[L;U,P] = LU(A)
[L,U,P] = LU(A) computes the LU factorization of matrix A.
The factorization is A = P**L**U, where P is a permutation
matrix, L is lower triangular with unit diagonal and U is
upper triangular.
[lu] = LU(A) leaves the factors L and U packed in one matrix.
[lu,p] = LU(A) returns also the pivoting info vector p.
(Notice that this p is related to the permutation matrix P
but is not the same. You need this form of LU if you want to
use LUbacksubst later on.)
See also:
LUbacksubst,
linsolve,
inv,
chol,
SVD.
Error codes:
1: Singular matrix (==> zero in U's diagonal)
-1: Input arg not an array
-2: Input arg not a rank-2 array (matrix)
[x] = LUbacksubst(lu,p,b)
LUbacksubst(lu,p,b) solves the linear system A**x == b,
where A has been previously LU-decomposed using LU:
[lu,p] = LU(A).
See also:
linsolve,
LU.
Error codes:
-1: First input arg is not an array
-2: First input arg is not a square matrix
-3: Second input arg is not an integer vector
-4: Third input arg is not an array
-5: Third input arg is not a vector or matrix
-6: Incompatible dimensions in first/third args
-7: Singular matrix
-8: Incompatible dimensions in first/second args
-9: Internal error
[C] = matprod(A,B; Aflag,Bflag)
matprod(A,B) returns the matrix product of A and B.
If at least one of A and B is scalar, matprod(A,B) is the
same as their ordinary product A*B. If both A and B
are arrays, their "inner" dimensions must agree.
That is, the last dimension of A must equal the first
dimension of B.
You can abbreviate matprod(A,B) as A**B.
Optional args: matprod(A,B,aflag,bflag) can be used to
transpose or Hermitian-conjugate the factors before the
product. 'n' means no operation, 't' means transpose and
'h' means Hermitian conjugate. For example,
matprod(A,B,'h') = A'**B = herm(A)**B
matprod(A,B,'n','t') = A**B.' = A**transpose(B)
Normally you need not use matprod explicitly, but you
can use the operator **, which is internally translated
to matprod. Hermitian conjugates and transposes in
connection with ** produce the corresponding 'h' and
't' options in matprod. For example,
A'**B generates matprod(A,B,'h')
A.'**B' generates matprod(A,B,'t','h')
A**B.' generates matprod(A,B,'n','t')
and so on. The runtime is optimal for all these operations.
See also:
inv.
Error codes:
-1: Inner dimensions do not agree
-2: Resulting array would have too high rank
-3: Third arg not one of 'n', 't', 'h'
-4: Fourth arg not one of 'n', 't', 'h'
[U;S,V] = SVD(A)
[U,S,V] = SVD(A) computes the singular value
decomposition of matrix A: A = U**S**V'.
U and V are unitary and S is diagonal.
SVD(A) as such returns the vector of singular values.
See also:
LU,
chol.
Error codes:
1: No convergence
-1: Input arg not an array
-2: Input array is not a matrix
-3: Internal error
-4: Two output arg case not supported (must be 1 or 3)