#include <lsqr.hh>
Inheritance diagram for RVLUmin::LSQRStep< Scalar >:
Public Member Functions | |
LSQRStep (LinearOp< Scalar > const &_A, Vector< Scalar > &_x, Vector< Scalar > const &_b, atype &_rnorm, atype &_nrnorm) | |
void | run () |
Run a single step of Paige-Saunders - notation as on p. | |
~LSQRStep () |
8(1), 1982, pp. 43-71.
On construction, internal workspace allocated and initialized. Each step updates internal state of LSQRStep object. Since solution vector, residual norm, and normal residual norm are stored as mutable references to external objects, these external objects are updated as well.
IMPORTANT NOTE: this version of the algorithm assumes that the solution vector reference (internal data member x) refers to a zero vector on initialization. To accommodate nontrivial initial guess, modify the right-hand-side vector (argument _b) externally.
Solution vector (x), iteration count, residual norm, and normal residual (gradient) norm are all references to external objects, which may be monitored by appropriate terminators to build a LoopAlg out of this Algorithm.
See LSQRAlg for description of a fully functional algorithm class, combining this step with a Terminator to make a LoopAlg.
Definition at line 77 of file lsqr.hh.
RVLUmin::LSQRStep< Scalar >::LSQRStep | ( | LinearOp< Scalar > const & | _A, | |
Vector< Scalar > & | _x, | |||
Vector< Scalar > const & | _b, | |||
atype & | _rnorm, | |||
atype & | _nrnorm | |||
) |
RVLUmin::LSQRStep< Scalar >::~LSQRStep | ( | ) |
void RVLUmin::LSQRStep< Scalar >::run | ( | ) | [virtual] |
Run a single step of Paige-Saunders - notation as on p.
8 of TOMS paper
Implements RVLAlg::Algorithm.