Logo ROOT   6.12/07
Reference Guide
multidimfit.C
Go to the documentation of this file.
1 /// \file
2 /// \ingroup tutorial_fit
3 /// \notebook -nodraw
4 /// Multi-Dimensional Parametrisation and Fitting
5 ///
6 /// \macro_output
7 /// \macro_code
8 ///
9 /// \authors Rene Brun, Christian Holm Christensen
10 
11 #include "Riostream.h"
12 #include "TROOT.h"
13 #include "TApplication.h"
14 #include "TCanvas.h"
15 #include "TH1.h"
16 #include "TSystem.h"
17 #include "TBrowser.h"
18 #include "TFile.h"
19 #include "TRandom.h"
20 #include "TMultiDimFit.h"
21 #include "TVectorD.h"
22 #include "TMath.h"
23 
24 
25 //____________________________________________________________________
26 void makeData(Double_t* x, Double_t& d, Double_t& e)
27 {
28  // Make data points
29  Double_t upp[5] = { 10, 10, 10, 10, 1 };
30  Double_t low[5] = { 0, 0, 0, 0, .1 };
31  for (int i = 0; i < 4; i++)
32  x[i] = (upp[i] - low[i]) * gRandom->Rndm() + low[i];
33 
34  d = x[0] * TMath::Sqrt(x[1] * x[1] + x[2] * x[2] + x[3] * x[3]);
35 
36  e = gRandom->Gaus(upp[4],low[4]);
37 }
38 
39 //____________________________________________________________________
40 int CompareResults(TMultiDimFit *fit, bool doFit)
41 {
42  //Compare results with reference run
43 
44 
45  // the right coefficients (before fit)
46  double GoodCoeffsNoFit[] = {
47  -4.37056,
48  43.1468,
49  13.432,
50  13.4632,
51  13.3964,
52  13.328,
53  13.3016,
54  13.3519,
55  4.49724,
56  4.63876,
57  4.89036,
58  -3.69982,
59  -3.98618,
60  -3.86195,
61  4.36054,
62  -4.02597,
63  4.57037,
64  4.69845,
65  2.83819,
66  -3.48855,
67  -3.97612
68  };
69 
70  // the right coefficients (after fit)
71  double GoodCoeffs[] = {
72  -4.399,
73  43.15,
74  13.41,
75  13.49,
76  13.4,
77  13.23,
78  13.34,
79  13.29,
80  4.523,
81  4.659,
82  4.948,
83  -4.026,
84  -4.045,
85  -3.939,
86  4.421,
87  -4.006,
88  4.626,
89  4.378,
90  3.516,
91  -4.111,
92  -3.823,
93  };
94 
95  // Good Powers
96  int GoodPower[] = {
97  1, 1, 1, 1,
98  2, 1, 1, 1,
99  1, 1, 1, 2,
100  1, 1, 2, 1,
101  1, 2, 1, 1,
102  2, 2, 1, 1,
103  2, 1, 1, 2,
104  2, 1, 2, 1,
105  1, 1, 1, 3,
106  1, 3, 1, 1,
107  1, 1, 5, 1,
108  1, 1, 2, 2,
109  1, 2, 1, 2,
110  1, 2, 2, 1,
111  2, 1, 1, 3,
112  2, 2, 1, 2,
113  2, 1, 3, 1,
114  2, 3, 1, 1,
115  1, 2, 2, 2,
116  2, 1, 2, 2,
117  2, 2, 2, 1
118  };
119 
120  Int_t nc = fit->GetNCoefficients();
121  Int_t nv = fit->GetNVariables();
122  const Int_t *powers = fit->GetPowers();
123  const Int_t *pindex = fit->GetPowerIndex();
124  if (nc != 21) return 1;
125  const TVectorD *coeffs = fit->GetCoefficients();
126  int k = 0;
127  for (Int_t i=0;i<nc;i++) {
128  if (doFit) {
129  if (!TMath::AreEqualRel((*coeffs)[i],GoodCoeffs[i],1e-3)) return 2;
130  }
131  else {
132  if (TMath::Abs((*coeffs)[i] - GoodCoeffsNoFit[i]) > 5e-5) return 2;
133  }
134  for (Int_t j=0;j<nv;j++) {
135  if (powers[pindex[i]*nv+j] != GoodPower[k]) return 3;
136  k++;
137  }
138  }
139 
140  // now test the result of the generated function
141  gROOT->ProcessLine(".L MDF.C");
142 
143  Double_t refMDF = (doFit) ? 43.95 : 43.98;
144  // this does not work in CLing since the function is not defined
145  //Double_t x[] = {5,5,5,5};
146  //Double_t rMDF = MDF(x);
147  //LM: need to return the address of the result since it is casted to a long (this should not be in a tutorial !)
148  Long_t iret = gROOT->ProcessLine(" Double_t xvalues[] = {5,5,5,5}; double result=MDF(xvalues); &result;");
149  Double_t rMDF = * ( (Double_t*)iret);
150  //printf("%f\n",rMDF);
151  if (TMath::Abs(rMDF -refMDF) > 1e-2) return 4;
152  return 0;
153 }
154 
155 //____________________________________________________________________
156 Int_t multidimfit(bool doFit = true)
157 {
158 
159  cout << "*************************************************" << endl;
160  cout << "* Multidimensional Fit *" << endl;
161  cout << "* *" << endl;
162  cout << "* By Christian Holm <cholm@nbi.dk> 14/10/00 *" << endl;
163  cout << "*************************************************" << endl;
164  cout << endl;
165 
166  // Initialize global TRannom object.
167  gRandom = new TRandom();
168 
169  // Open output file
170  TFile* output = new TFile("mdf.root", "RECREATE");
171 
172  // Global data parameters
173  Int_t nVars = 4;
174  Int_t nData = 500;
175  Double_t x[4];
176 
177  // make fit object and set parameters on it.
178  TMultiDimFit* fit = new TMultiDimFit(nVars, TMultiDimFit::kMonomials,"v");
179 
180  Int_t mPowers[] = { 6 , 6, 6, 6 };
181  fit->SetMaxPowers(mPowers);
182  fit->SetMaxFunctions(1000);
183  fit->SetMaxStudy(1000);
184  fit->SetMaxTerms(30);
185  fit->SetPowerLimit(1);
186  fit->SetMinAngle(10);
187  fit->SetMaxAngle(10);
188  fit->SetMinRelativeError(.01);
189 
190  // variables to hold the temporary input data
191  Double_t d;
192  Double_t e;
193 
194  // Print out the start parameters
195  fit->Print("p");
196 
197  printf("======================================\n");
198 
199  // Create training sample
200  Int_t i;
201  for (i = 0; i < nData ; i++) {
202 
203  // Make some data
204  makeData(x,d,e);
205 
206  // Add the row to the fit object
207  fit->AddRow(x,d,e);
208  }
209 
210  // Print out the statistics
211  fit->Print("s");
212 
213  // Book histograms
214  fit->MakeHistograms();
215 
216  // Find the parameterization
217  fit->FindParameterization();
218 
219  // Print coefficents
220  fit->Print("rc");
221 
222  // Get the min and max of variables from the training sample, used
223  // for cuts in test sample.
224  Double_t *xMax = new Double_t[nVars];
225  Double_t *xMin = new Double_t[nVars];
226  for (i = 0; i < nVars; i++) {
227  xMax[i] = (*fit->GetMaxVariables())(i);
228  xMin[i] = (*fit->GetMinVariables())(i);
229  }
230 
231  nData = fit->GetNCoefficients() * 100;
232  Int_t j;
233 
234  // Create test sample
235  for (i = 0; i < nData ; i++) {
236  // Make some data
237  makeData(x,d,e);
238 
239  for (j = 0; j < nVars; j++)
240  if (x[j] < xMin[j] || x[j] > xMax[j])
241  break;
242 
243  // If we get through the loop above, all variables are in range
244  if (j == nVars)
245  // Add the row to the fit object
246  fit->AddTestRow(x,d,e);
247  else
248  i--;
249  }
250  //delete gRandom;
251 
252  // Test the parameterizatio and coefficents using the test sample.
253  if (doFit)
254  fit->Fit("M");
255 
256  // Print result
257  fit->Print("fc v");
258 
259  // Write code to file
260  fit->MakeCode();
261 
262  // Write histograms to disk, and close file
263  output->Write();
264  output->Close();
265  delete output;
266 
267  // Compare results with reference run
268  Int_t compare = CompareResults(fit, doFit);
269  if (!compare) {
270  printf("\nmultidimfit .............................................. OK\n");
271  } else {
272  printf("\nmultidimfit .............................................. fails case %d\n",compare);
273  }
274 
275  // We're done
276  delete fit;
277  return compare;
278 }
virtual Double_t Gaus(Double_t mean=0, Double_t sigma=1)
Samples a random number from the standard Normal (Gaussian) Distribution with the given mean and sigm...
Definition: TRandom.cxx:256
virtual void Fit(Option_t *option="")
Try to fit the found parameterisation to the test sample.
A ROOT file is a suite of consecutive data records (TKey instances) with a well defined format...
Definition: TFile.h:46
TVectorT.
Definition: TMatrixTBase.h:77
void SetMaxPowers(const Int_t *powers)
Set the maximum power to be considered in the fit for each variable.
#define gROOT
Definition: TROOT.h:402
const Int_t * GetPowers() const
Definition: TMultiDimFit.h:166
int Int_t
Definition: RtypesCore.h:41
Short_t Abs(Short_t d)
Definition: TMathBase.h:108
const TVectorD * GetMinVariables() const
Definition: TMultiDimFit.h:160
Double_t x[n]
Definition: legend1.C:17
void SetPowerLimit(Double_t limit=1e-3)
Set the user parameter for the function selection.
This is the base class for the ROOT Random number generators.
Definition: TRandom.h:27
Int_t GetNCoefficients() const
Definition: TMultiDimFit.h:162
void SetMinAngle(Double_t angle=1)
Set the min angle (in degrees) between a new candidate function and the subspace spanned by the previ...
void SetMaxAngle(Double_t angle=0)
Set the max angle (in degrees) between the initial data vector to be fitted, and the new candidate fu...
const TVectorD * GetCoefficients() const
Definition: TMultiDimFit.h:142
virtual void AddRow(const Double_t *x, Double_t D, Double_t E=0)
Add a row consisting of fNVariables independent variables, the known, dependent quantity, and optionally, the square error in the dependent quantity, to the training sample to be used for the parameterization.
const TVectorD * GetMaxVariables() const
Definition: TMultiDimFit.h:154
void SetMaxTerms(Int_t terms)
Definition: TMultiDimFit.h:201
virtual Double_t Rndm()
Machine independent random number generator.
Definition: TRandom.cxx:533
Bool_t AreEqualRel(Double_t af, Double_t bf, Double_t relPrec)
Definition: TMath.h:322
Int_t GetNVariables() const
Definition: TMultiDimFit.h:161
virtual void Print(Option_t *option="ps") const
Print statistics etc.
virtual Int_t Write(const char *name=0, Int_t opt=0, Int_t bufsiz=0)
Write memory objects to this file.
Definition: TFile.cxx:2313
Multidimensional Fits in ROOT.
Definition: TMultiDimFit.h:15
virtual void FindParameterization(Option_t *option="")
Find the parameterization.
Int_t * GetPowerIndex() const
Definition: TMultiDimFit.h:164
R__EXTERN TRandom * gRandom
Definition: TRandom.h:62
virtual void AddTestRow(const Double_t *x, Double_t D, Double_t E=0)
Add a row consisting of fNVariables independent variables, the known, dependent quantity, and optionally, the square error in the dependent quantity, to the test sample to be used for the test of the parameterization.
virtual void MakeHistograms(Option_t *option="A")
Make histograms of the result of the analysis.
long Long_t
Definition: RtypesCore.h:50
double Double_t
Definition: RtypesCore.h:55
void SetMinRelativeError(Double_t error)
Set the acceptable relative error for when sum of square residuals is considered minimized.
you should not use this method at all Int_t Int_t Double_t Double_t Double_t e
Definition: TRolke.cxx:630
virtual void MakeCode(const char *functionName="MDF", Option_t *option="")
Generate the file <filename> with .C appended if argument doesn&#39;t end in .cxx or .C.
Double_t Sqrt(Double_t x)
Definition: TMath.h:590
void SetMaxStudy(Int_t n)
Definition: TMultiDimFit.h:200
void SetMaxFunctions(Int_t n)
Definition: TMultiDimFit.h:198
virtual void Close(Option_t *option="")
Close a file.
Definition: TFile.cxx:916