A bounds-constarined R implementation of a truncated Newton method for minimization of nonlinear functions subject to bounds (box) constraints.
tnbc(x, fgfun, lower, upper, trace=0, ...)A list with components:
The best set of parameters found.
The value of the objective at the best set of parameters found.
The gradient of the objective at the best set of parameters found.
An integer indicating the situation on termination. 0
indicates that the method believes it has succeeded; 2 that
more than maxfun (default 150*n, where there are n parameters);
3 if the line search appears to have failed (which may not be serious);
and -1 if there appears to be an error in the input parameters.
A number giving a measure of how many conjugate gradient solutions were used during the minimization process.
A numeric vector of starting estimates.
A function that returns the value of the objective at
the supplied set of parameters par using auxiliary data in
.... The gradient is returned as attribute "gradient".
The first argument of fgfun must be par.
A vector of lower bounds on the parameters.
A vector of upper bounds on the parameters.
Set >0 to cause intermediate output to allow progress to be followed.
Further arguments to be passed to fn.
Function fgfun must return a numeric value in list item f
and a numeric vector in list item g.
Stephen G. Nash (1984) "Newton-type minimization via the Lanczos method", SIAM J Numerical Analysis, vol. 21, no. 4, pages 770-788.
For Matlab code, see http://www.netlib.org/opt/tn