portelli/LatAnalyze
1
0
Fork 0
mirror of https://github.com/aportelli/LatAnalyze.git synced 2025-06-18 07:17:05 +01:00
Code Releases Activity

simplification of the Minuit minimiser interface, (successful) test of the new fit interface

Browse Source
This commit is contained in:
Antonin Portelli
2016-03-23 17:08:25 +00:00
parent ee8ed05b81
commit 9b9c86cf72
5 changed files with 168 additions and 185 deletions
Download Patch File Download Diff File Expand all files Collapse all files

42
examples/exFit.cpp
View File

@ -1,6 +1,7 @@
#include <iostream>
#include <cmath>
#include <LatAnalyze/CompiledModel.hpp>
#include <LatAnalyze/Io.hpp>
#include <LatAnalyze/MinuitMinimizer.hpp>
#include <LatAnalyze/RandGen.hpp>
#include <LatAnalyze/XYStatData.hpp>
@ -8,39 +9,50 @@
using namespace std;
using namespace Latan;
const Index nPoint = 30;
const double xErr = .2, yErr = .1;
const double exactPar[2] = {0.5,5.0}, dx = 10.0/static_cast<double>(nPoint);
const Index nPoint1 = 5, nPoint2 = 5;
const double xErr = .1, yErr = .1;
const double exactPar[2] = {0.5,5.};
const double dx1 = 10.0/static_cast<double>(nPoint1);
const double dx2 = 5.0/static_cast<double>(nPoint2);
int main(void)
{
// generate fake data
XYStatData data;
RandGen rg;
double x1_k, x2_k;
double xBuf[2];
DoubleModel f([](const double *x, const double *p)
{return p[1]*exp(-x[0]*p[0]);}, 1, 2);
{return p[1]*exp(-x[0]*p[0])+x[1];}, 2, 2);
data.addXDim("x", nPoint);
data.addXDim("x", nPoint1);
data.addXDim("off", nPoint2);
data.addYDim("y");
for (Index k = 0; k < nPoint; ++k)
for (Index i1 = 0; i1 < nPoint1; ++i1)
{
x1_k = rg.gaussian(k*dx, xErr);
x2_k = rg.gaussian(k*dx, xErr);
data.x(k) = x1_k;
data.y(k) = rg.gaussian(f(&x2_k, exactPar), yErr);
printf("% 8e % 8e % 8e % 8e\n", data.x(k), data.y(k), xErr, yErr);
xBuf[0] = i1*dx1;
data.x(i1, 0) = rg.gaussian(xBuf[0], xErr);
for (Index i2 = 0; i2 < nPoint2; ++i2)
{
xBuf[1] = i2*dx2;
data.x(i2, 1) = xBuf[1];
data.y(data.dataIndex(i1, i2)) = rg.gaussian(f(xBuf, exactPar),
yErr);
printf("% 8e % 8e % 8e % 8e % 8e\n", data.x(i1, 0), xErr,
data.x(i2, 1), data.y(i1), yErr);
}
}
cout << endl;
data.setXError(0, DVec::Constant(nPoint, xErr));
data.setYError(0, DVec::Constant(nPoint, yErr));
data.setXError(0, DVec::Constant(data.getXSize(0), xErr));
data.assumeXExact(true, 1);
data.setYError(0, DVec::Constant(data.getYSize(), yErr));
cout << data << endl;
// fit
DVec init = DVec::Constant(2, 0.5);
DVec init = DVec::Constant(2, 0.1);
FitResult p;
MinuitMinimizer minimizer;
minimizer.setVerbosity(Minimizer::Verbosity::Normal);
p = data.fit(minimizer, init, f);
cout << "a= " << p(0) << " b= " << p(1) << endl;
cout << "chi^2/ndof= " << p.getChi2PerDof() << endl;

42
examples/exFitSample.cpp
View File

@ -8,41 +8,49 @@
using namespace std;
using namespace Latan;
const Index nPoint = 30;
const Index nPoint1 = 10, nPoint2 = 10;
const Index nSample = 1000;
const double xErr = .2, yErr = .1;
const double exactPar[2] = {0.5,5.0}, dx = 10.0/static_cast<double>(nPoint);
const double xErr = .1, yErr = .1;
const double exactPar[2] = {0.5,5.};
const double dx1 = 10.0/static_cast<double>(nPoint1);
const double dx2 = 5.0/static_cast<double>(nPoint2);
int main(void)
{
// generate fake data
DMatSample x(nSample, nPoint, 1), y(nSample, nPoint, 1);
XYSampleData data(nSample);
RandGen rg;
double x1_k, x2_k;
DoubleModel f([](const double *t, const double *p)
{return p[1]*exp(-t[0]*p[0]);}, 1, 2);
RandGen rg;
double xBuf[2];
DoubleModel f([](const double *x, const double *p)
{return p[1]*exp(-x[0]*p[0])+x[1];}, 2, 2);
data.addXDim("x", nPoint);
data.addXDim("x", nPoint1);
data.addXDim("off", nPoint2);
data.addYDim("y");
FOR_STAT_ARRAY(x, s)
for (Index s = central; s < nSample; ++s)
{
for (Index k = 0; k < nPoint; ++k)
for (Index i1 = 0; i1 < nPoint1; ++i1)
{
x1_k = rg.gaussian(k*dx, xErr);
x2_k = rg.gaussian(k*dx, xErr);
data.x(k)[s] = x1_k;
data.y(k)[s] = rg.gaussian(f(&x2_k, exactPar), yErr);
xBuf[0] = i1*dx1;
data.x(i1, 0)[s] = rg.gaussian(xBuf[0], xErr);
for (Index i2 = 0; i2 < nPoint2; ++i2)
{
xBuf[1] = i2*dx2;
data.x(i2, 1)[s] = xBuf[1];
data.y(data.dataIndex(i1, i2))[s] =
rg.gaussian(f(xBuf, exactPar), yErr);
}
}
}
data.assumeXExact(true, 1);
cout << data << endl;
// fit
DVec init = DVec::Constant(2, 0.5);
DVec init = DVec::Constant(2, 0.1);
DMat err;
SampleFitResult p;
MinuitMinimizer minimizer;
p = data.fit(minimizer, init, f);
err = p.variance().cwiseSqrt();
cout << "a= " << p[central](0) << " +/- " << err(0) << endl;