// @(#)root/mathmore:$Id: GSLNLSMinimizer.h 26866 2008-12-12 10:50:07Z moneta $
// Author: L. Moneta Wed Dec 20 17:16:32 2006

/**********************************************************************
 *                                                                    *
 * Copyright (c) 2006  LCG ROOT Math Team, CERN/PH-SFT                *
 *                                                                    *
 * This library is free software; you can redistribute it and/or      *
 * modify it under the terms of the GNU General Public License        *
 * as published by the Free Software Foundation; either version 2     *
 * of the License, or (at your option) any later version.             *
 *                                                                    *
 * This library is distributed in the hope that it will be useful,    *
 * but WITHOUT ANY WARRANTY; without even the implied warranty of     *
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU   *
 * General Public License for more details.                           *
 *                                                                    *
 * You should have received a copy of the GNU General Public License  *
 * along with this library (see file COPYING); if not, write          *
 * to the Free Software Foundation, Inc., 59 Temple Place, Suite      *
 * 330, Boston, MA 02111-1307 USA, or contact the author.             *
 *                                                                    *
 **********************************************************************/

// Header file for class GSLNLSMinimizer

#ifndef ROOT_Math_GSLNLSMinimizer
#define ROOT_Math_GSLNLSMinimizer



#ifndef ROOT_Math_Minimizer
#include "Math/Minimizer.h"
#endif


#ifndef ROOT_Math_IFunctionfwd
#include "Math/IFunctionfwd.h"
#endif

#ifndef ROOT_Math_IParamFunctionfwd
#include "Math/IParamFunctionfwd.h"
#endif


#include "Math/FitMethodFunction.h"
//#include "Math/Derivator.h"



namespace ROOT { 

   namespace Math { 

      class GSLMultiFit; 


//________________________________________________________________________________
/** 
    LSResidualFunc class description. 
    Internal class used for accessing the residuals of the Least Square function
    and their derivates which are estimated numerically using GSL numerical derivation. 
    The class contains a pointer to the fit method function and an index specifying 
    the i-th residual and wraps it in a multi-dim gradient function interface
    ROOT::Math::IGradientFunctionMultiDim. 
    The class is used by ROOT::Math::GSLNLSMinimizer (GSL non linear least square fitter)

    @ingroup MultiMin
*/
class LSResidualFunc : public IMultiGradFunction { 
public: 

   //default ctor (required by CINT) 
   LSResidualFunc() : fIndex(0), fChi2(0)
   {}


   LSResidualFunc(const ROOT::Math::FitMethodFunction & func, unsigned int i) : 
      fIndex(i), 
      fChi2(&func), 
      fX2(std::vector<double>(func.NDim() ) )
   {}


   // copy ctor
   LSResidualFunc(const LSResidualFunc & rhs) :
      IMultiGenFunction(), 
      IMultiGradFunction() 
   { 
      operator=(rhs);
   } 

   // assignment
   LSResidualFunc & operator= (const LSResidualFunc & rhs) 
   { 
      fIndex = rhs.fIndex;
      fChi2 = rhs.fChi2; 
      fX2 = rhs.fX2;
      return *this;
   } 

   IMultiGenFunction * Clone() const { 
      return new LSResidualFunc(*fChi2,fIndex); 
   }

   unsigned int NDim() const { return fChi2->NDim(); }

   void Gradient( const double * x, double * g) const { 
      unsigned int n = NDim(); 
      std::copy(x,x+n,fX2.begin());
      const double kEps = 1.0E-4;
      double f0 = DoEval(x); 
      for (unsigned int i = 0; i < n; ++i) { 
         fX2[i] += kEps;
         g[i] =  ( DoEval(&fX2.front()) - f0 )/kEps;
         fX2[i] = x[i];
      }
   } 

private: 

   double DoEval (const double * x) const { 
      return fChi2->DataElement(x, fIndex);
   }
   
   double DoDerivative(const double * x, unsigned int icoord) const { 
      //return  ROOT::Math::Derivator::Eval(*this, x, icoord, 1E-8);
      std::copy(x,x+NDim(),fX2.begin());
      const double kEps = 1.0E-4;
      fX2[icoord] += kEps;
      return ( DoEval(&fX2.front()) - DoEval(x) )/kEps;
   }

   unsigned int fIndex; 
   const ROOT::Math::FitMethodFunction * fChi2; 
   mutable std::vector<double> fX2;  // cached vector
};


//_____________________________________________________________________________________________________
/** 
   GSLNLSMinimizer class for Non Linear Least Square fitting
   It Uses the Levemberg-Marquardt algorithm from 
   <A HREF="http://www.gnu.org/software/gsl/manual/html_node/Nonlinear-Least_002dSquares-Fitting.html">
   GSL Non Linear Least Square fitting</A>.

   @ingroup MultiMin
*/ 
class GSLNLSMinimizer : public  ROOT::Math::Minimizer {

public: 

   /** 
      Default constructor
   */ 
   GSLNLSMinimizer (int type = 0); 

   /** 
      Destructor (no operations)
   */ 
   ~GSLNLSMinimizer ();  

private:
   // usually copying is non trivial, so we make this unaccessible

   /** 
      Copy constructor
   */ 
   GSLNLSMinimizer(const GSLNLSMinimizer &) : ROOT::Math::Minimizer() {} 

   /** 
      Assignment operator
   */ 
   GSLNLSMinimizer & operator = (const GSLNLSMinimizer & rhs)  {
      if (this == &rhs) return *this;  // time saving self-test
      return *this;
   }

public: 

   /// set the function to minimize
   virtual void SetFunction(const ROOT::Math::IMultiGenFunction & func); 

   /// set gradient the function to minimize
   virtual void SetFunction(const ROOT::Math::IMultiGradFunction & func); 

   /// set free variable 
   virtual bool SetVariable(unsigned int ivar, const std::string & name, double val, double step); 

#ifdef LATER
   /// set lower limit variable  (override if minimizer supports them )
   virtual bool SetLowerLimitedVariable(unsigned int  ivar , const std::string & name , double val , double step , double lower );
   /// set upper limit variable (override if minimizer supports them )
   virtual bool SetUpperLimitedVariable(unsigned int ivar , const std::string & name , double val , double step , double upper ); 
   /// set upper/lower limited variable (override if minimizer supports them )
   virtual bool SetLimitedVariable(unsigned int ivar , const std::string & name , double val , double step , double /* lower */, double /* upper */); 
   /// set fixed variable (override if minimizer supports them )
   virtual bool SetFixedVariable(unsigned int /* ivar */, const std::string & /* name */, double /* val */);  
#endif

   /// set the value of an existing variable 
   virtual bool SetVariableValue(unsigned int ivar, double val );
   /// set the values of all existing variables (array must be dimensioned to the size of existing parameters)
   virtual bool SetVariableValues(const double * x);

   /// method to perform the minimization
   virtual  bool Minimize(); 

   /// return minimum function value
   virtual double MinValue() const { return fMinVal; } 

   /// return expected distance reached from the minimum
   virtual double Edm() const { return 0; } // not impl. }

   /// return  pointer to X values at the minimum 
   virtual const double *  X() const { return &fValues.front(); } 

   /// return pointer to gradient values at the minimum 
   virtual const double *  MinGradient() const; 

   /// number of function calls to reach the minimum 
   virtual unsigned int NCalls() const { return 0; } // not yet ipl.  

   /// this is <= Function().NDim() which is the total 
   /// number of variables (free+ constrained ones) 
   virtual unsigned int NDim() const { return fDim; }   

   /// number of free variables (real dimension of the problem) 
   /// this is <= Function().NDim() which is the total 
   virtual unsigned int NFree() const { return fDim; }  

   /// minimizer provides error and error matrix
   virtual bool ProvidesError() const { return true; } 

   /// return errors at the minimum 
   virtual const double * Errors() const { return &fErrors.front(); }
//  { 
//       static std::vector<double> err; 
//       err.resize(fDim);
//       return &err.front(); 
//    }

   /** return covariance matrices elements 
       if the variable is fixed the matrix is zero
       The ordering of the variables is the same as in errors
   */ 
   virtual double CovMatrix(unsigned int , unsigned int ) const;

   /// minos error for variable i, return false if Minos failed
   virtual bool GetMinosError(unsigned int , double & /* errLow */ , double & /* errUp */ ) { return false; }

   /// return reference to the objective function
   ///virtual const ROOT::Math::IGenFunction & Function() const; 


protected: 

private: 
   
   // dimension of the function to be minimized 
   unsigned int fDim; 
   // number of fit points (residuals)
   unsigned int fSize; 

   ROOT::Math::GSLMultiFit * fGSLMultiFit;        // pointer to GSL multi fit solver 
   const ROOT::Math::IMultiGenFunction * fObjFunc;      // pointer to Least square function
   
   double fMinVal;                                // minimum function value
   double fLSTolerance;                           // Line Search Tolerance
   std::vector<double> fValues;           
   std::vector<double> fErrors;
   const double * fCovMatrix;         // pointer to cov matrix (stored in fGSLMultiFit)
   std::vector<double> fSteps;
   std::vector<std::string> fNames;
   std::vector<LSResidualFunc> fResiduals;   //! transient Vector of the residual functions

}; 

   } // end namespace Math

} // end namespace ROOT


#endif /* ROOT_Math_GSLNLSMinimizer */