vnl_lsqr.h

Go to the documentation of this file.

#ifndef vnl_lsqr_h_
#define vnl_lsqr_h_
//:
// \file
// \brief Linear least squares
// \author David Capel, capes@robots, July 2000
//    vnl_lsqr implements an algorithm for large, sparse linear systems and
//    sparse, linear least squares. It is a wrapper for the LSQR algorithm
//    of Paige and Saunders (ACM TOMS 583). The sparse system is encapsulated
//    by a vnl_linear_system.
//
//     

//   Modifications
//     000705 capes@robots initial version.
//     4/4/01 LSB (Manchester) Documentation tidied
//
//-----------------------------------------------------------------------------

#include <vnl/vnl_vector.h>
#include <vnl/vnl_matrix.h>
#include <vnl/vnl_linear_system.h>


class vnl_lsqr {
public:
  vnl_lsqr(vnl_linear_system& ls) :
    ls_(&ls), max_iter_(ls.get_number_of_unknowns()) {}

  ~vnl_lsqr();

  void set_max_iterations(int max_iter) { max_iter_ = max_iter; }

  //: Perform the minimization starting from x=0 and putting the result into x.
  // Return code may be translated with translate_return_code().
  int minimize(vnl_vector<double>& x);

  int get_number_of_iterations() const { return num_iter_; }

  //: Pontificate about the outcome of the last minimization.
  void diagnose_outcome(vcl_ostream& os) const;

  static void translate_return_code(vcl_ostream& os, int return_code);

protected:
  vnl_linear_system* ls_;
  int max_iter_;
  int num_iter_;
  double resid_norm_estimate_;
  double result_norm_estimate_;
  double A_condition_estimate_;
  double result_norm_;
  int return_code_;

  static int aprod_(int* mode, int* m, int* n, double* x, 
    double* y, int* leniw, int* lenrw, int* iw, double* rw );

  friend class vnl_lsqr_Activate;

};

#endif // vnl_lsqr_h_

---