EMT Pt 38 nanoparticle

SystemEMT molecular dynamics at 300 K

Atoms38

Runs50

Cap10000 epochs

The main metric is Mean epochs. Scoring rules.

Charts

Success rate on EMT Pt 38 nanoparticle, in rank order. The rate is the share of runs that finished. 6 ranked rows did not record a success rate and are not drawn. PDF.

Dolan-Moré profile for Minimizers on EMT Pt 38 nanoparticle. The curve is the fraction of starts whose cost is at most tau times the best finite cost in this family. Tau is 1, 2, 4, 8, and 16. An unsolved start stays in the denominator and does not cross a finite tau. Dolan, E. D. and Moré, J. J., Math. Program. 91:201-213, 2002. PDF.

Force calls on each recorded start for EMT Pt 38 nanoparticle. The box is the quartiles of those counts. A blue point is a success and a vermillion cross is a recorded failure (50). The horizontal position is the recorded count. The vertical offset only keeps marks from covering one another. PDF.

Performance profile. Higher means more starts solved within that cost ratio.

• n2p2 (2.3.0) · Kalman energy
• n2p2 (2.3.0) · Kalman force

Cost distribution of 2 ranked rows. Logarithmic bins, from 25.02 to 60000.

Performance profile SVG Cost and success SVG

Performance profile. Select the plot for the full SVG.

Cost and success. Select the plot for the full SVG.

A shorter bar is a lower cost on a log scale shared by the codes of this problem.

Leaderboard

A shorter bar is a lower cost on the log scale for that problem. Select a column heading to sort.

Rank	Code	Best	Median	Success	PAR-2
1	n2p2	25.02	30012.5	100.0%	25.02

The tables below list every ranked row, by method family and overall.

These archived rows rank on the value each one published, which leaves out any failure. The failure rule has no score for them, for the reason given with each.

• Pyamff (8037eb98) · pyamff-lbfgs10-7-fcoeff0 (the problem states no cap)

Minimizers

Minimizers. One row is one code, one version, and one method. A lower cost is better. PAR-2 counts a finished run as the calls it used and a failed run as twice the cap. An archived row shows the average it published, and its PAR-2 applies that rule to the failures the entry records. Hessians is the mean analytic Hessian count. FD calls is that count times one force call per coordinate.

Rank	Row	Cost	Unit	Hessians	FD calls	Runs
1	n2p2 (2.3.0) · Kalman energy	25.02	loss-gradient evaluations	0	0	50
2	n2p2 (2.3.0) · Kalman force	60000	loss-gradient evaluations	0	0	50

Overall

One row is one code, one version, and one method. A lower cost is better. PAR-2 counts a finished run as the calls it used and a failed run as twice the cap. An archived row shows the average it published, and its PAR-2 applies that rule to the failures the entry records. Hessians is the mean analytic Hessian count. FD calls is that count times one force call per coordinate.

Row	Family	Family rank	Unit	PAR-2	Published	Median calls	Mean calls	Hessians	FD calls	Failures	Runs
n2p2 (2.3.0) · Kalman energy	Minimizers	1	loss-gradient evaluations	25.02		14	25.02	0	0	crash 0, divergence 0, cap 0, wrong_stationary 0	50
n2p2 (2.3.0) · Kalman force	Minimizers	2	loss-gradient evaluations	60000		30000	30000	0	0	crash 0, divergence 0, cap 50, wrong_stationary 0	50

A row that counts another unit is ranked in its own table after the overall table.

Ranked in epochs

These rows count epochs. The tables above count loss-gradient evaluations, so these rows are ranked only among themselves.

The archived mean_epoch is the benchmark website's Mean Epochs: mlrf0_optb.py -s averages the last epoch in each model's pyamff.log. optbench/n2p2_ml.py charges one loss-gradient evaluation before each Kalman epoch. No source equates a pyamff epoch with that count.

One row is one code, one version, and one method. A lower cost is better. PAR-2 counts a finished run as the calls it used and a failed run as twice the cap. An archived row shows the average it published, and its PAR-2 applies that rule to the failures the entry records. Hessians is the mean analytic Hessian count. FD calls is that count times one force call per coordinate.

Row	Family	Family rank	Unit	PAR-2	Published	Median calls	Mean calls	Hessians	FD calls	Failures	Runs
Pyamff (8037eb98) · pyamff-rprop10-7-fcoeff0	Minimizers	1	epochs		97.58			0	0	0 failed	50
Pyamff (8037eb98) · pyamff-rprop10-7-fcoeff0_1	Minimizers	2	epochs		1022.86			0	0	0 failed	50
Pyamff (8037eb98) · pyamff-adam10-7-fcoeff0	Minimizers	3	epochs		1103.24			0	0	0 failed	50
Pyamff (8037eb98) · pyamff-adam10-7-fcoeff0_1	Minimizers	4	epochs		1708.8			0	0	0 failed	50
Pyamff (8037eb98) · pyamff-lbfgs10-7-fcoeff0	Minimizers	5	epochs		1791.8			0	0	1 failed, class not recorded	50
Pyamff (8037eb98) · pyamff-lbfgs10-7-fcoeff0_1	Minimizers	6	epochs		1943.34			0	0	0 failed	50