forum.alglib.net
http://forum.alglib.net/

Calling ALGLIB from MATLAB mex-files
http://forum.alglib.net/viewtopic.php?f=2&t=4369
Page 1 of 1

Author:  matthiaskaldorf [ Thu Feb 11, 2021 2:25 pm ]
Post subject:  Calling ALGLIB from MATLAB mex-files

Hi everybody!

I am currently trying to add several ALGLIB-routines into a MATLAB program via mex-files. Linking to ALGLIB seems to be more evolved than adding header-only libraries such as Eigen. Specifically, I was trying to wrap the Levenberg-Marquardt-example file into a mex-file that can subsequently be compiled and called from MATLAB. Compiling this via the 'mex'-command however does produce a huge array of errors. To be sure, I copied and compiled all ALGLIB cpp files into the same directory from Visual Studio 2019 and also verified that the Levenberg-Marquardt example works when called from Visual Studio. Does anybody have experience with these issues or can provide me some info on how to link ALGLIB in this case?

Thanks in Advance
Matthias

Code:
#include <mex.h>

#include "stdafx.h"
#include <stdlib.h>
#include <stdio.h>
#include <math.h>
#include "optimization.h"

using namespace alglib;
void  function1_fvec(const real_1d_array &x, real_1d_array &fi, void *ptr)
{
    //
    // this callback calculates
    // f0(x0,x1) = 100*(x0+3)^4,
    // f1(x0,x1) = (x1-3)^4
    //
    fi[0] = 10*pow(x[0]+3,2);
    fi[1] = pow(x[1]-3,2);
}

void mexFunction(int nlhs, mxArray *plhs[], int nrhs, const mxArray *prhs[])
{
    //
    // This example demonstrates minimization of F(x0,x1) = f0^2+f1^2, where
    //
    //     f0(x0,x1) = 10*(x0+3)^2
    //     f1(x0,x1) = (x1-3)^2
    //
    // using "V" mode of the Levenberg-Marquardt optimizer.
    //
    // Optimization algorithm uses:
    // * function vector f[] = {f1,f2}
    //
    // No other information (Jacobian, gradient, etc.) is needed.
    //
    real_1d_array x = "[0,0]";
    real_1d_array s = "[1,1]";
    double epsx = 0.0000000001;
    ae_int_t maxits = 0;
    minlmstate state;
    minlmreport rep;

    //
    // Create optimizer, tell it to:
    // * use numerical differentiation with step equal to 0.0001
    // * use unit scale for all variables (s is a unit vector)
    // * stop after short enough step (less than epsx)
    //
    minlmcreatev(2, x, 0.0001, state);
    minlmsetcond(state, epsx, maxits);
    minlmsetscale(state, s);

    //
    // Optimize
    //
    alglib::minlmoptimize(state, function1_fvec);

    //
    // Test optimization results
    //
    // NOTE: because we use numerical differentiation, we do not
    //       verify Jacobian correctness - it is always "correct".
    //       However, if you switch to analytic gradient, consider
    //       checking it with OptGuard (see other examples).
    //
    minlmresults(state, x, rep);
    mexPrintf("%s\n", x.tostring(2).c_str()); // EXPECTED: [-3,+3]
   
    // point to elements of output list
    double *OUTPUT;
   
    // allocate output list
    plhs[0] = mxCreateDoubleMatrix(1, 1, mxREAL);
    OUTPUT  = mxGetPr(prhs[0]);

    OUTPUT[0] = 1;
}

Page 1 of 1 All times are UTC
Powered by phpBB © 2000, 2002, 2005, 2007 phpBB Group
http://www.phpbb.com/