@c Copyright (C) 1996 John W. Eaton
@c This is part of the Octave manual.
@c For copying conditions, see the file gpl.texi.

@node Optimization, Quadrature, Differential Equations, Top
@chapter Optimization

@menu
* Quadratic Programming::       
* Nonlinear Programming::       
* Linear Least Squares::        
@end menu

@c @cindex linear programming
@cindex quadratic programming
@cindex nonlinear programming
@cindex optimization
@cindex LP
@cindex QP
@cindex NLP

@node Quadratic Programming, Nonlinear Programming, Optimization, Optimization
@section Quadratic Programming

@ftable @code
@item qpsol

@example
[x, obj, info, lambda]
   = qpsol (x, H, c, lb, ub, lb, A, ub)
@end example
Solve quadratic programs using Gill and Murray's QPSOL.  Because QPSOL
is not freely redistributable, this function is only available if you
have obtained your own copy of QPSOL.  @xref{Installation}.
@end ftable

@findex qpsol_options
Tolerances and other options for @code{qpsol} may be specified using the
function @code{qpsol_options}.

@node Nonlinear Programming, Linear Least Squares, Quadratic Programming, Optimization
@section Nonlinear Programming

@ignore
@ftable @code
@item fsqp

@example
fsqp ()
@end example

@findex fsqp_options
Tolerances and other options for @code{fsqp} may be specified using the
function @code{fsqp_options}.

Sorry, this hasn't been implemented yet.
@end ignore

@ftable @code
@item npsol

@example
[x, obj, info, lambda]
   = npsol (x, 'phi', lb, ub, lb, A, ub, lb, 'g', ub)
@end example
Solve nonlinear programs using Gill and Murray's NPSOL.  Because NPSOL
is not freely redistributable, this function is only available if you
have obtained your own copy of NPSOL.  @xref{Installation}.

The second argument is a string containing the name of the objective
function to call.  The objective function must be of the form

@example
y = phi (x)
@end example

@noindent
where x is a vector and y is a scalar.
@end ftable

@findex npsol_options
Tolerances and other options for @code{npsol} may be specified using the
function @code{npsol_options}.

@node Linear Least Squares,  , Nonlinear Programming, Optimization
@section Linear Least Squares

@ftable @code
@item gls (@var{Y}, @var{X}, @var{O})
Generalized least squares (GLS) estimation for the multivariate model

@example
Y = X * B + E,  mean(E) = 0,  cov(vec(E)) = (s^2)*O
@end example

@noindent
with

@example
Y an T x p matrix
X an T x k matrix
B an k x p matrix
E an T x p matrix
O an Tp x Tp matrix
@end example

@noindent
Each row of Y and X is an observation and each column a variable.

Returns BETA, v, and, R, where BETA is the GLS estimator for B, v is the
GLS estimator for s^2, and R = Y - X*BETA is the matrix of GLS residuals.

@item ols (@var{Y}, @var{X})
Ordinary Least Squares (OLS) estimation for the multivariate model

@example
Y = X*B + E,  mean (E) = 0,  cov (vec (E)) = kron (S, I)
@end example

@noindent
with

@example
Y an T x p matrix
X an T x k matrix
B an k x p matrix
E an T x p matrix
@end example

@noindent
Each row of Y and X is an observation and each column a variable.

Returns BETA, SIGMA, and R, where BETA is the OLS estimator for B, i.e.

@example
BETA = pinv(X)*Y,
@end example

@noindent
where pinv(X) denotes the pseudoinverse of X, SIGMA is the OLS estimator
for the matrix S, i.e.

@example
SIGMA = (Y - X*BETA)'*(Y - X*BETA) / (T - rank(X))
@end example

and R = Y - X*BETA is the matrix of OLS residuals.
@end ftable