=pod
=begin html
<link rel="stylesheet" href="podstyle.css" type="text/css" />
=end html
=head1 NAME
lmmin - Levenberg-Marquardt least-squares minimization
=head1 SYNOPSIS
B<#include <lmmin.h>>
B<void lmmin( const int> I<n_par>B<, double *>I<par>B<, const int> I<m_dat>B<,
constS< >void *>I<data>B<,
void *>I<evaluate>B<(
constS< >double *>I<par>B<, const int >I<m_dat>B<,
constS< >void *>I<data>B<, double *>I<fvec>B<, int *>I<userbreak>B<),
constS< >lm_control_struct *>I<control>B<,
lm_status_struct *>I<status>B< );>
B<extern const lm_control_struct lm_control_double;>
B<extern const lm_control_struct lm_control_float;>
B<extern const char *lm_infmsg[];>
B<extern const char *lm_shortmsg[];>
=head1 DESCRIPTION
B<lmmin()> determines a vector I<par> that minimizes the sum of squared elements of a vector I<fvec> that is computed by a user-supplied function I<evaluate>().
On success, I<par> represents a local minimum, not necessarily a global one; it may depend on its starting value.
For applications in curve fitting, the wrapper function B<lmcurve(3)> offers a simplified API.
The Levenberg-Marquardt minimization starts with a steepest-descent exploration of the parameter space, and achieves rapid convergence by crossing over into the Newton-Gauss method.
Function arguments:
=over
=item I<n_par>
Number of free variables.
Length of parameter vector I<par>.
=item I<par>
Parameter vector.
On input, it must contain a reasonable guess.
On output, it contains the solution found to minimize ||I<fvec>||.
=item I<m_dat>
Length of vector I<fvec>.
Must statisfy I<n_par> <= I<m_dat>.
=item I<data>
This pointer is ignored by the fit algorithm,
except for appearing as an argument in all calls to the user-supplied
routine I<evaluate>.
=item I<evaluate>
Pointer to a user-supplied function that computes I<m_dat> elements of vector I<fvec> for a given parameter vector I<par>.
If I<evaluate> return with *I<userbreak> set to a negative value, B<lmmin()> will interrupt the fitting and terminate.
=item I<control>
Parameter collection for tuning the fit procedure.
In most cases, the default &I<lm_control_double> is adequate.
If I<f> is only computed with single-precision accuracy,
I<&lm_control_float> should be used.
See also below, NOTES on initializing parameter records.
I<control> has the following members (for more details, see the source file I<lmstruct.h>):
=over
=item B<double> I<control.ftol>
Relative error desired in the sum of squares.
Recommended setting: somewhat above machine precision; less if I<fvec> is computed with reduced accuracy.
=item B<double> I<control.xtol>
Relative error between last two approximations.
Recommended setting: as I<ftol>.
=item B<double> I<control.gtol>
A measure for degeneracy.
Recommended setting: as I<ftol>.
=item B<double> I<control.epsilon>
Step used to calculate the Jacobian.
Recommended setting: as I<ftol>, but definitely less than the accuracy of I<fvec>.
=item B<double> I<control.stepbound>
Initial bound to steps in the outer loop, generally between 0.01 and 100; recommended value is 100.
=item B<int> I<control.patience>
Used to set the maximum number of function evaluations to patience*n_par.
=item B<int> I<control.scale_diag>
Logical switch (0 or 1).
If 1, then scale parameters to their initial value.
This is the recommended setting.
=item B<FILE*> I<control.msgfile>
Progress messages will be written to this file.
Typically I<stdout> or I<stderr>.
The value I<NULL> will be interpreted as I<stdout>.
=item B<int> I<control.verbosity>
If nonzero, some progress information from within the LM algorithm
is written to control.stream.
=item B<int> I<control.n_maxpri>
-1, or maximum number of parameters to print.
=item B<int> I<control.m_maxpri>
-1, or maximum number of residuals to print.
=back
=item I<status>
A record used to return information about the minimization process:
=over
=item B<double> I<status.fnorm>
Norm of the vector I<fvec>;
=item B<int> I<status.nfev>
Actual number of iterations;
=item B<int> I<status.outcome>
Status of minimization;
for the corresponding text message, print I<lm_infmsg>B<[>I<status.outcome>B<]>;
for a short code, print I<lm_shortmsg>B<[>I<status.outcome>B<]>.
=item B<int> I<status.userbreak>
Set when termination has been forced by the user-supplied routine I<evaluate>.
=back
=back
=head1 NOTES
=head2 Initializing parameter records.
The parameter record I<control> should always be initialized
from supplied default records:
lm_control_struct control = lm_control_double; /* or _float */
After this, parameters may be overwritten:
control.patience = 500; /* allow more iterations */
control.verbosity = 15; /* for verbose monitoring */
An application written this way is guaranteed to work even if new parameters
are added to I<lm_control_struct>.
Conversely, addition of parameters is not considered an API change; it may happen without increment of the major version number.
=head1 EXAMPLES
=head2 Fitting a surface
Fit a data set y(t) by a function f(t;p) where t is a two-dimensional vector:
#include "lmmin.h"
#include <stdio.h>
/* fit model: a plane p0 + p1*tx + p2*tz */
double f( double tx, double tz, const double *p )
{
return p[0] + p[1]*tx + p[2]*tz;
}
/* data structure to transmit data arays and fit model */
typedef struct {
double *tx, *tz;
double *y;
double (*f)( double tx, double tz, const double *p );
} data_struct;
/* function evaluation, determination of residues */
void evaluate_surface( const double *par, int m_dat,
const void *data, double *fvec, int *userbreak )
{
/* for readability, explicit type conversion */
data_struct *D;
D = (data_struct*)data;
int i;
for ( i = 0; i < m_dat; i++ )
fvec[i] = D->y[i] - D->f( D->tx[i], D->tz[i], par );
}
int main()
{
/* parameter vector */
int n_par = 3; /* number of parameters in model function f */
double par[3] = { -1, 0, 1 }; /* arbitrary starting value */
/* data points */
int m_dat = 4;
double tx[4] = { -1, -1, 1, 1 };
double tz[4] = { -1, 1, -1, 1 };
double y[4] = { 0, 1, 1, 2 };
data_struct data = { tx, tz, y, f };
/* auxiliary parameters */
lm_status_struct status;
lm_control_struct control = lm_control_double;
control.verbosity = 3;
/* perform the fit */
printf( "Fitting:\n" );
lmmin( n_par, par, m_dat, (const void*) &data, evaluate_surface,
&control, &status );
/* print results */
printf( "\nResults:\n" );
printf( "status after %d function evaluations:\n %s\n",
status.nfev, lm_infmsg[status.outcome] );
printf("obtained parameters:\n");
int i;
for ( i=0; i<n_par; ++i )
printf(" par[%i] = %12g\n", i, par[i]);
printf("obtained norm:\n %12g\n", status.fnorm );
printf("fitting data as follows:\n");
double ff;
for ( i=0; i<m_dat; ++i ){
ff = f(tx[i], tz[i], par);
printf( " t[%2d]=%12g,%12g y=%12g fit=%12g residue=%12g\n",
i, tx[i], tz[i], y[i], ff, y[i] - ff );
}
return 0;
}
=head2 More examples
For more examples, see the homepage and directories demo/ and test/ in the source distribution.
=head1 COPYING
Copyright (C):
1980-1999 University of Chicago
2004-2015 Joachim Wuttke, Forschungszentrum Juelich GmbH
Software: FreeBSD License
Documentation: Creative Commons Attribution Share Alike
=head1 SEE ALSO
=begin html
<a href="http://apps.jcns.fz-juelich.de/man/lmcurve.html"><b>lmcurve</b>(3)</a>
=end html
=begin man
\fBlmcurve\fR(3)
.PP
=end man
Homepage: http://apps.jcns.fz-juelich.de/lmfit
=head1 BUGS
Please send bug reports and suggestions to the author <j.wuttke@fz-juelich.de>.