forum.alglib.net

ALGLIB forum
It is currently Tue Jul 07, 2020 6:32 pm

All times are UTC


Forum rules


1. This forum can be used for discussion of both ALGLIB-related and general numerical analysis questions
2. This forum is English-only - postings in other languages will be removed.



Post new topic Reply to topic  [ 1 post ] 
Author Message
 Post subject: minlmoptimize doesnt do anything! Help please!
PostPosted: Thu Jun 11, 2020 4:56 pm 
Offline

Joined: Thu Jun 04, 2020 3:29 pm
Posts: 2
Hello,

I am trying to fit a curve of the form:
f(x)=c0+c1*tanh(c2*(x-c3))
to 300 points of 2d data e.g. "x" and "y" coordinates, in my case "x" is index and "y" is some given known numerical value.
I need this to be done fast and so after solving this with lsfit I tried to implement using minlm.
Since minlm minimizes a function of sum of squares form i.e.,
F=f1^2+f2^2+...fm^2
I defined my functions to be the errors of the data for a given guess of coefficients i.e.
fi=c0+c1*tanh(c2*(i-c3))-data(i)
and analytically computed the jacobian matrix (300X4).
Unfortunately, minlmresults seems to do nothing and I cant figure out why. Please help me, I have nowhere else to turn.
I post my code below and attach my data in a txt file. "esf" is the 300 points of data I mentions:

using System;
using System.IO;
using System.Text;
using System.Linq;

namespace minlm_d_vj_example
{
class Program
{
public static int Main(string[] args)
{
string esf_string = File.ReadAllText(@"...4_1_1_esf.txt"); //Read data from file
string[] esf_string_split = esf_string.Split('\n');
double[] esf = new double[esf_string_split.Length-1];

double[] ci = new double[4]; //In this block I initialize the coefficients with a clever guess
ci[0] = (esf[0] + esf[^1]) / 2;
ci[1] = (esf[0] - esf[^1]) / 2;
int d = 0;
double dThreshold = ci[0];
for (; (d < esf.Length) && (esf[d] > dThreshold); d++) ;
ci[3] = d;
int di = 2;
double f1 = esf[d - di];
double f2 = esf[d + di];
ci[2] = -(f1 - f2) / (2 * di) / ci[1];

double[] s = new double[] { 1, 1, 1, 1};
double epsx = 0.0000000001;
int maxits = 0;

alglib.minlmstate state;
alglib.minlmcreatevj(esf.Length, ci, out state);
alglib.minlmsetcond(state, epsx, maxits);
alglib.minlmsetscale(state, s);
//alglib.minlmoptguardgradient(state, 0.001);
alglib.minlmoptimize(state, function1_fvec, function1_jac, null, null);
alglib.minlmreport rep;
alglib.minlmresults(state, out ci, out rep);
System.Console.WriteLine("{0}", alglib.ap.format(ci, 2));
alglib.optguardreport ogrep;
alglib.minlmoptguardresults(state, out ogrep);
System.Console.WriteLine("{0}", ogrep.badgradsuspected);
System.Console.ReadLine();
return 0;

void function1_fvec(double[] c, double[] fi, object obj)
{
for (int i = 0; i < esf.Length; i++)
{
fi[i] = c[0] + c[1] * Math.Tanh(c[2] * (i - c[3])) - esf[i];
}
}
void function1_jac(double[] c, double[] fi, double[,] jac, object obj)
{
for (int i = 0; i < esf.Length; i++)
{
fi[i] = c[0] + c[1] * Math.Tanh(c[2] * (i - c[3])) - esf[i];
double k = i - c[3];
jac[i, 0] = 1; //Here I used some algebra to avoid unnecessary computations. OptGuard gives True but I can't find any problems with the jacobian.
jac[i, 1] = Math.Tanh(c[2] * k);
jac[i, 2] = c[1] * (1 - jac[i, 1] * jac[i, 1]) * k;
jac[i, 3] = jac[i, 2] * (-c[2] / k);
}
}
}
}
}


Attachments:
File comment: This files contains the data to be fitted as a function of index.
4_1_1_esf.txt [2.64 KiB]
Downloaded 3 times
Top
 Profile  
 
Display posts from previous:  Sort by  
Post new topic Reply to topic  [ 1 post ] 

All times are UTC


Who is online

Users browsing this forum: No registered users and 1 guest


You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot post attachments in this forum

Search for:
Jump to:  
cron
Powered by phpBB © 2000, 2002, 2005, 2007 phpBB Group