Optimizer tests

This page shows benchmarks of optimizations done with ASE’s different optimizers. Note that the iteration number (steps) is not the same as the number of force evaluations. This is because some of the optimizers uses internal line searches or similar.

The most important performance characteristics of an optimizer is the total optimization time. Different optimizers may perform the same number of steps, but along a different path, so the time spent on calculation of energy/forces may be different due to different convergence of the self-consistent field.

Test systems

These are the test systems (systems.db):

test-name	description
H2	Hydrogen molecule
Cu16	Shaken bulk copper
Cu2	Distorted Cu(111) surface
CAu8O	CO on Au(111) surface
C5H12	Pentane molecule
CCu8	C/Cu(100)
Al13	Alumninum cluster

EMT calculations

Calculation done with EMT. Number of steps:

optimizer	H2	Cu16	Cu2	Au8CO	Cu8C	Al13
GPMin	4	6	11	15	4	17
BFGSLineSearch	7	7	22	25	7	32
GoodOldQuasiNewton	12	10	18	28	10	33
BFGS	9	15	25	41	12	55
SciPyFminBFGS	5	16	29	44	15	60
SciPyFminCG	11	18	35	68	18	55
PreconLBFGS	11	16	57	43	18	63
FIRE	30	40	64	52	31	48
LBFGSLineSearch	8	17		48	14	58
LBFGS		15	25	41	12	55
MDMin		10	20	69	11	39
PreconFIRE	25				91

GPAW-LCAO calculations

Parameters:

GPAW(mode='lcao',
     basis='dzp',
     kpts={'density': 2.0})

Absolute time relative to fastest optimizer:

optimizer	H2	Cu16	Cu2	Au8CO	C5H12	Cu8C	Al13
GPMin	1.1	1.0	1.0	1.0	1.7	1.3	1.0
BFGSLineSearch	1.3	1.1	1.9	1.0	3.0	1.0	1.2
GoodOldQuasiNewton	1.8	2.1	2.1	1.2	1.1	1.1	1.3
SciPyFminBFGS	1.0	2.2	2.4	1.2	1.0	1.5	1.9
BFGS	1.6	2.2	2.1	1.2	1.0	1.5	1.9
LBFGSLineSearch	1.3	2.7	2.3	1.3	1.2	1.5	2.4
SciPyFminCG	1.8	2.5	2.6	1.6	2.0	2.0	1.9
PreconLBFGS	1.6	2.4	5.3	1.8	1.1	3.5	2.3
LBFGS		2.2	2.1	2.0	1.1	1.6	2.0
MDMin	1.3	1.7	1.9		3.9	1.5	1.6
FIRE	2.3	5.9	4.2		2.9	3.0	1.9
PreconFIRE	4.8				5.0