search

numbbo / coco

Numerical Black-Box Optimization Benchmarking Framework
https://numbbo.github.io/coco
Other
260 stars 86 forks source link

Readability of Scaling figures #1797

Open nikohansen opened 6 years ago

nikohansen commented 6 years ago
screen shot 2018-06-05 at 10 35 42

The readability is compromised by the max-evaluations light symbols. They should appear only if not a single run was successful and hence no average could be computed.

EDIT: they should appear only if the average (aRT) is above the max-evaluation or missing (because we divide by the success ratio).

brockho commented 6 years ago

Also no number appears when comparing BFGS-P-StPt NIPOPa! 2009/NEWUOA! lmm! NELDERDOERR! RANDOMSEARCH! and MCS!. Not clear though whether this was a feature of the single-algorithm plots only or whether this already appeared in the past in the multiple-algorithms plots as well.