Previous topic

mapp.md.min_cg.run

Next topic

mapp.md.min_lbfgs.H_dof

mapp.md.min_lbfgs

class mapp.md.min_lbfgs(m=2,e_tol=1.0e-8,H_dof=[[False],[False,False],[False,False,False]],affine=False,max_dx=1.0,ls=mapp.md.ls_bt())

L-BFGS minimization algorithm

Parameters:

m : int

Maximum number of vectors to be stored in memory

e_tol : double

Energy tolerance criterion for stopping minimization

H_dof : symm<bool[dim][dim]>

Unitcell degrees of freedom during minimization, here dim is the dimension of simulation

affine : bool

If set to True atomic displacements would be affine

max_dx : double

Maximum displacement of any atom in one step of minimization

ls : mapp.ls

Line search method

Notes

Limited memory Broyden–Fletcher–Goldfarb–Shanno (L-BFGS) algorithm for minimization, see [1].

References

[1]Jorge Nocedal and Stephen J. Wright. Numerical optimization. Springer series in operations research and financial engineering. New York : Springer, c2006., 2006. ISBN 978-0-387-30303-1.

Attributes

H_dof (symm<bool[dim][dim]>) unitcell DOFs
affine (bool) affine transformation switch
e_tol (double) tolerance
export (mapp.md.export) export object
ls (mapp.ls) line search algorithm
m (int) Maximum No. of vectors
max_dx (double) mximum displacemant
ntally (int) thermodynamic tallying period

Methods

run(atoms,max_nsteps) Execute minimization