Logo ROOT   6.13/01
Reference Guide
MinimTransformFunction.cxx
Go to the documentation of this file.
1 // @(#)root/mathmore:$Id$
2 // Author: L. Moneta June 2009
3 
4 /**********************************************************************
5  * *
6  * Copyright (c) 2006 LCG ROOT Math Team, CERN/PH-SFT *
7  * *
8  * *
9  **********************************************************************/
10 
11 // Implementation file for class MinimTransformFunction
12 
14 #include "Math/IFunctionfwd.h"
15 
16 //#include <iostream>
17 #include <cmath>
18 #include <cassert>
19 
20 namespace ROOT {
21 
22  namespace Math {
23 
24 MinimTransformFunction::MinimTransformFunction ( const IMultiGradFunction * f, const std::vector<EMinimVariableType> & types,
25  const std::vector<double> & values,
26  const std::map<unsigned int, std::pair<double, double> > & bounds) :
27  fX( values ),
28  fFunc(f)
29 {
30  // constructor of the class from a pointer to the function (which is managed)
31  // vector specifying the variable types (free, bounded or fixed, defined in enum EMinimVariableTypes )
32  // variable values (used for the fixed ones) and a map with the bounds (for the bounded variables)
33 
34  unsigned int ntot = NTot(); // NTot is fFunc->NDim()
35  assert ( types.size() == ntot );
36  fVariables.reserve(ntot);
37  fIndex.reserve(ntot);
38  for (unsigned int i = 0; i < ntot; ++i ) {
39  if (types[i] == kFix )
40  fVariables.push_back( MinimTransformVariable( values[i]) );
41  else {
42  fIndex.push_back(i);
43 
44  if ( types[i] == kDefault)
45  fVariables.push_back( MinimTransformVariable() );
46  else {
47  std::map<unsigned int, std::pair<double,double> >::const_iterator itr = bounds.find(i);
48  assert ( itr != bounds.end() );
49  double low = itr->second.first;
50  double up = itr->second.second;
51  if (types[i] == kBounds )
52  fVariables.push_back( MinimTransformVariable( low, up, new SinVariableTransformation()));
53  else if (types[i] == kLowBound)
55  else if (types[i] == kUpBound)
57  }
58  }
59  }
60 }
61 
62 
63 void MinimTransformFunction::Transformation( const double * x, double * xext) const {
64  // transform from internal to external
65 
66  unsigned int nfree = fIndex.size();
67 
68 // std::cout << "Transform: internal ";
69 // for (int i = 0; i < nfree; ++i) std::cout << x[i] << " ";
70 // std::cout << "\t\t";
71 
72  for (unsigned int i = 0; i < nfree; ++i ) {
73  unsigned int extIndex = fIndex[i];
74  const MinimTransformVariable & var = fVariables[ extIndex ];
75  if (var.IsLimited() )
76  xext[ extIndex ] = var.InternalToExternal( x[i] );
77  else
78  xext[ extIndex ] = x[i];
79  }
80 
81 // std::cout << "Transform: external ";
82 // for (int i = 0; i < fX.size(); ++i) std::cout << fX[i] << " ";
83 // std::cout << "\n";
84 
85 }
86 
87 void MinimTransformFunction::InvTransformation(const double * xExt, double * xInt) const {
88  // inverse function transformation (external -> internal)
89  for (unsigned int i = 0; i < NDim(); ++i ) {
90  unsigned int extIndex = fIndex[i];
91  const MinimTransformVariable & var = fVariables[ extIndex ];
92  assert ( !var.IsFixed() );
93  if (var.IsLimited() )
94  xInt[ i ] = var.ExternalToInternal( xExt[extIndex] );
95  else
96  xInt[ i ] = xExt[extIndex];
97  }
98 }
99 
100 void MinimTransformFunction::InvStepTransformation(const double * x, const double * sExt, double * sInt) const {
101  // inverse function transformation for steps (external -> internal)
102  for (unsigned int i = 0; i < NDim(); ++i ) {
103  unsigned int extIndex = fIndex[i];
104  const MinimTransformVariable & var = fVariables[ extIndex ];
105  assert ( !var.IsFixed() );
106  if (var.IsLimited() ) {
107  // bound variables
108  double x2 = x[extIndex] + sExt[extIndex];
109  if (var.HasUpperBound() && x2 >= var.UpperBound() )
110  x2 = x[extIndex] - sExt[extIndex];
111  // transform x and x2
112  double xint = var.ExternalToInternal ( x[extIndex] );
113  double x2int = var.ExternalToInternal( x2 );
114  sInt[i] = std::abs( x2int - xint);
115  }
116  else
117  sInt[ i ] = sExt[extIndex];
118  }
119 }
120 
121 void MinimTransformFunction::GradientTransformation(const double * x, const double *gExt, double * gInt) const {
122  //transform gradient vector (external -> internal) at internal point x
123  unsigned int nfree = fIndex.size();
124  for (unsigned int i = 0; i < nfree; ++i ) {
125  unsigned int extIndex = fIndex[i];
126  const MinimTransformVariable & var = fVariables[ extIndex ];
127  assert (!var.IsFixed() );
128  if (var.IsLimited() )
129  gInt[i] = gExt[ extIndex ] * var.DerivativeIntToExt( x[i] );
130  else
131  gInt[i] = gExt[ extIndex ];
132  }
133 }
134 
135 
136 void MinimTransformFunction::MatrixTransformation(const double * x, const double *covInt, double * covExt) const {
137  //transform covariance matrix (internal -> external) at internal point x
138  // use row storages for matrices m(i,j) = rep[ i * dim + j]
139  // ignore fixed points
140  unsigned int nfree = fIndex.size();
141  unsigned int ntot = NTot();
142  for (unsigned int i = 0; i < nfree; ++i ) {
143  unsigned int iext = fIndex[i];
144  const MinimTransformVariable & ivar = fVariables[ iext ];
145  assert (!ivar.IsFixed());
146  double ddi = ( ivar.IsLimited() ) ? ivar.DerivativeIntToExt( x[i] ) : 1.0;
147  // loop on j variables for not fixed i variables (forget that matrix is symmetric) - could be optimized
148  for (unsigned int j = 0; j < nfree; ++j ) {
149  unsigned int jext = fIndex[j];
150  const MinimTransformVariable & jvar = fVariables[ jext ];
151  double ddj = ( jvar.IsLimited() ) ? jvar.DerivativeIntToExt( x[j] ) : 1.0;
152  assert (!jvar.IsFixed() );
153  covExt[ iext * ntot + jext] = ddi * ddj * covInt[ i * nfree + j];
154  }
155  }
156 }
157 
158 
159  } // end namespace Math
160 
161 } // end namespace ROOT
162 
Interface (abstract class) for multi-dimensional functions providing a gradient calculation.
Definition: IFunction.h:326
void MatrixTransformation(const double *x, const double *covInt, double *covExt) const
transform covariance matrix (internal -> external) at internal point x use row storages for matrices ...
Namespace for new ROOT classes and functions.
Definition: TFoamSampler.h:19
const double * Transformation(const double *x) const
transform from internal to external result is cached also inside the class
void InvTransformation(const double *xext, double *xint) const
inverse transformation (external -> internal)
unsigned int NDim() const
Retrieve the dimension of the function.
Sqrt Transformation class for dealing with upper bounded variables.
MinimTransformVariable class Contains meta information of the variables such as bounds, fix flags and deals with transformation of the variable The class does not contain the values and the step size (error) of the variable This is an internal class used by the MinimTransformFunction class.
* x
Deprecated and error prone model selection interface.
Definition: TRolke.cxx:630
std::vector< MinimTransformVariable > fVariables
void InvStepTransformation(const double *x, const double *sext, double *sint) const
inverse transformation for steps (external -> internal) at external point x
Namespace for new Math classes and functions.
MinimTransformFunction(const IMultiGradFunction *f, const std::vector< ROOT::Math::EMinimVariableType > &types, const std::vector< double > &values, const std::map< unsigned int, std::pair< double, double > > &bounds)
Constructor from a IMultiGradFunction interface (which is managed by the class) vector specifying the...
void GradientTransformation(const double *x, const double *gExt, double *gInt) const
transform gradient vector (external -> internal) at internal point x
Sin Transformation class for dealing with double bounded variables.
Sqrt Transformation class for dealing with lower bounded variables.