md_grad_optimizer< term_t > Class Template Reference

Gradient descent with parabolic minima search. More...

#include <gradopt.h>

Inheritance diagram for md_grad_optimizer< term_t >:

Collaboration diagram for md_grad_optimizer< term_t >:

Public Member Functions
virtual int	SetAccuracy (double max_dE, double max_dx, int maxconvit=2, double max_force=-1.)
	Set convergence creiteria: max_dE – maximal energy variance at convergence OR; maximal position variance at convergence, max_dx is normalized by total variable number OR maximal force norm max_force at convergence.

Protected Member Functions
int	tFunc (double &val, double t)
	function to calcualte value
int	ParabolicMin (FILE *f, int vout=1)
	makes linear search using tFunc, grdval

Detailed Description

template<class term_t>
class md_grad_optimizer< term_t >

Gradient descent with parabolic minima search.

Member Function Documentation

ParabolicMin()

template<class term_t >

int md_grad_optimizer< term_t >::ParabolicMin ( FILE *  f, int  vout = 1  )

protected

makes linear search using tFunc, grdval

Returns

SetAccuracy()

template<class term_t >

virtual int md_grad_optimizer< term_t >::SetAccuracy ( double  max_dE, double  max_dx, int  maxconvit = 2, double  max_force = -1.  )

inlinevirtual

Set convergence creiteria:
max_dE – maximal energy variance at convergence OR;
maximal position variance at convergence, max_dx is normalized by total variable number OR
maximal force norm max_force at convergence.

Negative parameters mean that the corresponding criteria are not checked. These conditions have to be repeated for at least maxconvit consecutive iterations.

Reimplemented from base_optimizer.

---

The documentation for this class was generated from the following file:

• gradopt.h