cgls.3.man

COOOL：CWP面向对象最优化库（CWP Object Oriented Optimization Library） COOOL是C++类的一个集合

CGLS(derived)     OPTIMIZATION ALGORITHMS      CGLS(derived)

                                                     Jun  1 15:17
NAME
    CGLS - Conjugate Gradient for Least Squares

SYNOPSIS
    #include <CGLS.hh>

    class LSConjugateGradient : public QuadraticOptima

        \fIPublic members\fP
            LSConjugateGradient(int, LinearForward*,
            Vector<double>*, int, double);
            LSConjugateGradient(int, LinearForward*,
            Vector<double>*, int, double, int);
            ~LSConjugateGradient();
            Model<double> 	optimizer(Model<double>&);
            Model<long> 	optimizer(Model<long>&) { return 0;}
            };
            #endif

        \fIProtected members\fP
            double		alpha;
            double		scale;
            double		scaleOld;
            Vector<double>*	search;
            Vector<double>*	modelError;
            Vector<double>*	dataError;
            void		getModelError(Vector<double>&);
            void		conjugateDirection();
            void		upDating();

DESCRIPTION
    LSConjugateGradient()
    The conjugate gradient implemented here should be used for the 
    solution of the normal equations A^T.A.x = A^T.y. It is coded 
    such that the product A^T.A is never performed, to avoid
    numerical instabilities and non sparse matrices. This procedure 
    comes straight from the classical paper "Methods of conjugate 
    gradients for solving linear systems:, 1952, NBS J. Research
    by Hesteness and Stiefel.

DESCRIPTION
    Constructors:
    LSConjugateGradient(int ???, LinearForward* ???, 
                        Vector<double>* ???, int ???, double ???);
    LSConjugateGradient(int ???, LinearForward* ???, Vector<double>* ???, 
                        int ???, double ???, int ???);
    ???...
    Method:
     Model<double> optimizer(Model<double>& model0);
      model0:  Initial model for the CGLS procedure.
      The optimum model is returned by the function.

CAVEATS
    Hmmm, there are no bugs in this code. The caveats that 
    I can think off are related to the facts that you are using 
    least squares, like sensitivity to outliers, non gaussian 
    statistics and so on :^). So...you are at your own risk.

DEFINED MACROS
    CGLS_HH

INCLUDED FILES
    "QuaOptima.hh"

SOURCE FILES
    CGLS.cc