Logo ROOT   6.18/05
Reference Guide
RMinimizer.cxx
Go to the documentation of this file.
1
2#include "TRInterface.h"
3#include "Math/RMinimizer.h"
4#include "Math/IFunction.h"
5#include <TVectorD.h>
7
8namespace ROOT {
9 namespace Math{
10
11 /// function wrapper for the function to be minimized
13 /// function wrapper for the gradient of the function to be minimized
15 /// integer for the number of function calls
16 int gNCalls = 0;
17
18 ///function to return the function values at point x
19 double minfunction(const std::vector<double> & x){
20 gNCalls++;
21 //return (*gFunction)(x.GetMatrixArray());
22 return (*gFunction)(x.data());
23 }
24 ///function to return the gradient values at point y
26 unsigned int size = y.GetNoElements();
27 const double * yy = y.GetMatrixArray();
28 double z[size];
30 TVectorD zz(size,z);
31 return zz;
32 }
33
34 /*Default constructor with option for the method of minimization, can be any of the following:
35 *
36 *"Nelder-Mead", "BFGS", "CG", "L-BFGS-B", "SANN", "Brent" (Brent only for 1D minimization)
37 */
39 fMethod=method;
40 if (fMethod.empty() || fMethod=="Migrad") fMethod="BFGS";
41 }
42
43 ///returns number of function calls
44 unsigned int RMinimizer::NCalls() const { return gNCalls; }
45
46 ///function for finding the minimum
48
49 //Set the functions
52
53 gNCalls = 0;
54
55 //pass functions and variables to R
57
58 r["minfunction"] = ROOT::R::TRFunctionExport(minfunction);
59 r["mingradfunction"] = ROOT::R::TRFunctionExport(mingradfunction);
60 r["method"] = fMethod.c_str();
61 std::vector<double> stepSizes(StepSizes(), StepSizes()+NDim());
62 std::vector<double> values(X(), X()+NDim());
63 r["ndim"] = NDim();
64 int ndim = NDim();
65 r["stepsizes"] = stepSizes;
66 r["initialparams"] = values;
67
68 //check if optimx is available
69 bool optimxloaded = FALSE;
70 r["optimxloaded"] = optimxloaded;
71 r.Execute("optimxloaded<-library(optimx,logical.return=TRUE)");
72 //int ibool = r.ParseEval("optimxloaded").ToScalar<Int_t>();
73 int ibool = r.Eval("optimxloaded");
74 if (ibool==1) optimxloaded=kTRUE;
75
76 //string for the command to be processed in R
77 TString cmd;
78
79 //optimx is available and loaded
80 if (optimxloaded==kTRUE) {
81 if (!gGradFunction) {
82 // not using gradient function
83 cmd = TString::Format("result <- optimx( initialparams, minfunction,method='%s',control = list(ndeps=stepsizes,maxit=%d,trace=%d,abstol=%e),hessian=TRUE)",fMethod.c_str(),MaxIterations(),PrintLevel(),Tolerance());
84 }
85 else {
86 // using user provided gradient
87 cmd = TString::Format("result <- optimx( initialparams, minfunction,mingradfunction, method='%s', control = list(ndeps=stepsizes,maxit=%d,trace=%d,abstol=%e),hessian=TRUE)",fMethod.c_str(),MaxIterations(),PrintLevel(),Tolerance());
88
89 }
90 }
91
92 //optimx is not available
93 else {
94 if (!gGradFunction) {
95 // not using gradient function
96 cmd = TString::Format("result <- optim( initialparams, minfunction,method='%s',control = list(ndeps=stepsizes,maxit=%d,trace=%d,abstol=%e),hessian=TRUE)",fMethod.c_str(),MaxIterations(),PrintLevel(),Tolerance());
97 }
98 else {
99 // using user provided gradient
100 cmd = TString::Format("result <- optim( initialparams, minfunction,mingradfunction, method='%s', control = list(ndeps=stepsizes,maxit=%d,trace=%d,abstol=%e),hessian=TRUE)",fMethod.c_str(),MaxIterations(),PrintLevel(),Tolerance());
101 }
102 }
103 //execute the minimization in R
104 std::cout << "Calling R with command " << cmd << std::endl;
105 r.Execute(cmd.Data());
106
107 //results with optimx
108 if (optimxloaded){
109 //get result from R
110 r.Execute("par<-coef(result)");
111 //get hessian matrix (in list form)
112 r.Execute("hess<-attr(result,\"details\")[,\"nhatend\"]");
113 //convert hess to a matrix
114 r.Execute("hess<-sapply(hess,function(x) x)");
115 //convert to square matrix
116 r.Execute("hess<-matrix(hess,c(ndim,ndim))");
117 //find covariant matrix from inverse of hess
118 r.Execute("cov<-solve(hess)");
119 //get errors from the sqrt of the diagonal of cov
120 r.Execute("errors<-sqrt(abs(diag(cov)))");
121 }
122
123 //results with optim
124 else {
125 r.Execute("par<-result$par");
126 r.Execute("hess<-result$hessian");
127 r.Execute("cov<-solve(hess)");
128 r.Execute("errors<-sqrt(abs(diag(cov)))");
129 }
130
131 //return the minimum to ROOT
132 //TVectorD vector = gR->ParseEval("par").ToVector<Double_t>();
133 std::vector<double> vectorPar = r["par"];
134
135 //get errors and matrices from R
136 // ROOT::R::TRObjectProxy p = gR->ParseEval("cov");
137 // TMatrixD cm = p.ToMatrix<Double_t>();
138 TMatrixD cm = r["cov"];
139 // p = gR->ParseEval("errors");
140 // TVectorD err = p.ToVector<Double_t>();
141 std::vector<double> err = r["errors"];
142 // p = gR->ParseEval("hess");
143 // TMatrixD hm = p.ToMatrix<Double_t>();
144 TMatrixD hm = r["hess"];
145
146 //set covariant and Hessian matrices and error vector
147 fCovMatrix.ResizeTo(ndim,ndim);
148 fHessMatrix.ResizeTo(ndim,ndim);
149 //fErrors.ResizeTo(ndim);
150 fCovMatrix = cm;
151 fErrors = err;
152 fHessMatrix = hm;
153
154 //get values and show minimum
155 const double *min=vectorPar.data();
156 SetFinalValues(min);
157 SetMinValue((*gFunction)(min));
158 std::cout<<"Value at minimum ="<<MinValue()<<std::endl;
159
160 return kTRUE;
161 }
162#ifdef LATER
163 //Returns the ith jth component of the covarient matrix
164 double RMinimizer::CovMatrix(unsigned int i, unsigned int j) const {
165 unsigned int ndim = NDim();
166 if (fCovMatrix==0) return 0;
167 if (i > ndim || j > ndim) return 0;
168 return fCovMatrix[i][j];
169 }
170 // //Returns the full parameter error vector
171 // TVectorD RMinimizer::RErrors() const {
172 // return fErrors;
173 // }
174 //Returns the ith jth component of the Hessian matrix
175 double RMinimizer::HessMatrix(unsigned int i, unsigned int j) const {
176 unsigned int ndim = NDim();
177 if (fHessMatrix==0) return 0;
178 if (i > ndim || j > ndim) return 0;
179 return fHessMatrix[i][j];
180 }
181#endif
182 } // end namespace MATH
183}
ROOT::R::TRInterface & r
Definition: Object.C:4
const Bool_t kTRUE
Definition: RtypesCore.h:87
const char Option_t
Definition: RtypesCore.h:62
#define FALSE
virtual double MinValue() const
return minimum function value
void SetMinValue(double val)
virtual const double * StepSizes() const
accessor methods
const ROOT::Math::IMultiGenFunction * ObjFunction() const
return pointer to used objective function
virtual unsigned int NDim() const
number of dimensions
virtual const double * X() const
return pointer to X values at the minimum
void SetFinalValues(const double *x)
const ROOT::Math::IMultiGradFunction * GradObjFunction() const
return pointer to used gradient object function (NULL if gradient is not supported)
Documentation for the abstract class IBaseFunctionMultiDim.
Definition: IFunction.h:62
Interface (abstract class) for multi-dimensional functions providing a gradient calculation.
Definition: IFunction.h:327
virtual void Gradient(const T *x, T *grad) const
Evaluate all the vector of function derivatives (gradient) at a point x.
Definition: IFunction.h:342
double Tolerance() const
absolute tolerance
Definition: Minimizer.h:420
unsigned int MaxIterations() const
max iterations
Definition: Minimizer.h:417
int PrintLevel() const
minimizer configuration parameters
Definition: Minimizer.h:411
double HessMatrix(unsigned int i, unsigned int j) const
Returns the ith jth component of the Hessian matrix.
virtual unsigned int NCalls() const
Returns the number of function calls.
Definition: RMinimizer.cxx:44
virtual double CovMatrix(unsigned int ivar, unsigned int jvar) const
return covariance matrices element for variables ivar,jvar if the variable is fixed the return value ...
Definition: RMinimizer.h:66
TMatrixD fCovMatrix
covariant matrix
Definition: RMinimizer.h:37
std::vector< double > fErrors
vector of parameter errors
Definition: RMinimizer.h:36
std::string fMethod
minimizer method to be used, must be of a type listed in R optim or optimx descriptions
Definition: RMinimizer.h:33
RMinimizer(Option_t *method)
Default constructor.
Definition: RMinimizer.cxx:38
TMatrixD fHessMatrix
Hessian matrix.
Definition: RMinimizer.h:38
virtual bool Minimize()
Function to find the minimum.
Definition: RMinimizer.cxx:47
This is a class to pass functions from ROOT to R.
ROOT R was implemented using the R Project library and the modules Rcpp and RInside
Definition: TRInterface.h:137
void Execute(const TString &code)
Method to eval R code.
Definition: TRInterface.cxx:82
static TRInterface & Instance()
static method to get an TRInterface instance reference
Int_t Eval(const TString &code, TRObject &ans)
Method to eval R code and you get the result in a reference to TRObject.
Definition: TRInterface.cxx:63
virtual TMatrixTBase< Element > & ResizeTo(Int_t nrows, Int_t ncols, Int_t=-1)
Set size of the matrix to nrows x ncols New dynamic elements are created, the overlapping part of the...
Definition: TMatrixT.cxx:1210
Basic string class.
Definition: TString.h:131
const char * Data() const
Definition: TString.h:364
static TString Format(const char *fmt,...)
Static method which formats a string using a printf style format descriptor and return a TString.
Definition: TString.cxx:2311
Double_t y[n]
Definition: legend1.C:17
Double_t x[n]
Definition: legend1.C:17
Namespace for new Math classes and functions.
const ROOT::Math::IMultiGenFunction * gFunction
function wrapper for the function to be minimized
Definition: RMinimizer.cxx:12
double minfunction(const std::vector< double > &x)
function to return the function values at point x
Definition: RMinimizer.cxx:19
TVectorD mingradfunction(TVectorD y)
function to return the gradient values at point y
Definition: RMinimizer.cxx:25
const ROOT::Math::IMultiGradFunction * gGradFunction
function wrapper for the gradient of the function to be minimized
Definition: RMinimizer.cxx:14
int gNCalls
integer for the number of function calls
Definition: RMinimizer.cxx:16
Namespace for new ROOT classes and functions.
Definition: StringConv.hxx:21
static constexpr double cm