Spaces:

---

calibrationcomp

/

calibration_benchmark

like 0

Sleeping

Issue

Currently, we do not provide explicit recommendations. We have some detailed "summary" information regarding overall scores across all targets/simulations. But in its current form it often combines objects that are not quite comparable, and may overly weight the minutia, with a large presentation of numbers, so it does not always help to give a clean recommendation.

I think something more helpful will be visual, and provide a simply "recommended algorithms" for problems of type "X"

Possible solutions

The user who wants a recommendation will often have a problem; so

• we should clearly state the dimensionality of the problem etc. we should not summarize over this information.
  They want a more binary recommendation (e.g. what is considered useable and what is not)
• we should focus on a generous criteria that can robustly categorize things into a binary (recommended/not recommended)
• we should present this in a simple visual on the home-page.

With this in mind the sketch below of a table with red-green (Rec/not-Rec) is