Logo Search packages:      
Sourcecode: quantlib version File versions  Download package


Go to the documentation of this file.
/* -*- mode: c++; tab-width: 4; indent-tabs-mode: nil; c-basic-offset: 4 -*- */

 Copyright (C) 2001, 2002, 2003 Nicolas Di Césaré
 Copyright (C) 2005, 2007 StatPro Italia srl

 This file is part of QuantLib, a free-software/open-source library
 for financial quantitative analysts and developers - http://quantlib.org/

 QuantLib is free software: you can redistribute it and/or modify it
 under the terms of the QuantLib license.  You should have received a
 copy of the license along with this program; if not, please email
 <quantlib-dev@lists.sf.net>. The license is also available online at

 This program 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 license for more details.

/*! \file leastsquare.hpp
    \brief Least square cost function

#ifndef quantlib_least_square_hpp
#define quantlib_least_square_hpp

#include <ql/math/optimization/problem.hpp>
#include <ql/math/optimization/conjugategradient.hpp>
#include <ql/math/matrix.hpp>

namespace QuantLib {

    class Constraint;
    class OptimizationMethod;

    //! Base class for least square problem
00038     class LeastSquareProblem {
        virtual ~LeastSquareProblem() {}
        //! size of the problem ie size of target vector
        virtual Size size() = 0;
        //! compute the target vector and the values of the function to fit
        virtual void targetAndValue(const Array& x,
                                    Array& target,
                                    Array& fct2fit) = 0;
        /*! compute the target vector, the values of the function to fit
            and the matrix of derivatives
        virtual void targetValueAndGradient(const Array& x,
                                            Matrix& grad_fct2fit,
                                            Array& target,
                                            Array& fct2fit) = 0;

    //! Cost function for least-square problems
    /*! Implements a cost function using the interface provided by
        the LeastSquareProblem class.
00060     class LeastSquareFunction : public CostFunction {
        //! Default constructor
00063         LeastSquareFunction(LeastSquareProblem& lsp) : lsp_(lsp) {}
        //! Destructor
00065         virtual ~LeastSquareFunction() {}

        //! compute value of the least square function
        virtual Real value(const Array& x) const;
        virtual Disposable<Array> values(const Array&) const;
        //! compute vector of derivatives of the least square function
        virtual void gradient(Array& grad_f, const Array& x) const;
        //! compute value and gradient of the least square function
        virtual Real valueAndGradient(Array& grad_f,
                                      const Array& x) const;
        //! least square problem
00077         LeastSquareProblem &lsp_;

    //! Non-linear least-square method.
    /*! Using a given optimization algorithm (default is conjugate

        \f[ min \{ r(x) : x in R^n \} \f]

        where \f$ r(x) = |f(x)|^2 \f$ is the Euclidean norm of \f$
        f(x) \f$ for some vector-valued function \f$ f \f$ from
        \f$ R^n \f$ to \f$ R^m \f$,
        \f[ f = (f_1, ..., f_m) \f]
        with \f$ f_i(x) = b_i - \phi(x,t_i) \f$ where \f$ b \f$ is the
        vector of target data and \f$ phi \f$ is a scalar function.

        Assuming the differentiability of \f$ f \f$, the gradient of
        \f$ r \f$ is defined by
        \f[ grad r(x) = f'(x)^t.f(x) \f]
00097     class NonLinearLeastSquare {
        //! Default constructor
        NonLinearLeastSquare(Constraint& c,
                             Real accuracy = 1e-4,
                             Size maxiter = 100);
        //! Default constructor
        NonLinearLeastSquare(Constraint& c,
                             Real accuracy,
                             Size maxiter,
                             boost::shared_ptr<OptimizationMethod> om);
        //! Destructor
00109         ~NonLinearLeastSquare() {}

        //! Solve least square problem using numerix solver
        Array& perform(LeastSquareProblem& lsProblem);

        void setInitialValue(const Array& initialValue) {
            initialValue_ = initialValue;

        //! return the results
00119         Array& results() { return results_; }

        //! return the least square residual norm
00122         Real residualNorm() { return resnorm_; }

        //! return last function value
00125         Real lastValue() { return bestAccuracy_; }

        //! return exit flag
00128         Integer exitFlag() { return exitFlag_; }

        //! return the performed number of iterations
00131         Integer iterationsNumber() { return nbIterations_; }
        //! solution vector
00134         Array results_, initialValue_;
        //! least square residual norm
00136         Real resnorm_;
        //! Exit flag of the optimization process
00138         Integer exitFlag_;
        //! required accuracy of the solver
00140         Real accuracy_, bestAccuracy_;
        //! maximum and real number of iterations
00142         Size maxIterations_, nbIterations_;
        //! Optimization method
00144         boost::shared_ptr<OptimizationMethod> om_;
        Constraint& c_;




Generated by  Doxygen 1.6.0   Back to index