search

BoostV / process-optimizer-frontend

A web frontend for ProcessOptimizer
BSD 3-Clause "New" or "Revised" License
10 stars 3 forks source link

159 number of initial experiments should be dynamically calculated #173

Closed langdal closed 2 years ago

dk-teknologisk-mon commented 1 year ago

The scaling of the number of initial experiments needs expert user input (and or benchmarking), since the present scaling of ´factors * 3´ is way too aggressive.

j-or commented 1 year ago

@dk-teknologisk-mon You can click the big 0/x number and edit "n initial points". Is this what you mean? If yes, perhaps we should add an edit button to make it more clear.

Screenshot 2022-11-16 at 15 45 58
dk-teknologisk-mon commented 1 year ago

No, I mean that the default suggestion (3 times the number of factors you enter) is way too much and actively scares users away. It is common that our users enter 5-8 factors, and in the latter case we right now tell them to go out and do 24 experiments before they can get anywhere. This makes them think "oh no, this is just like regular DOE where I have to do too many experiments" and so they give up before they even get started. If we base the number of suggested initial points on the factors, the scaling should probably just be one point per factor. This will be a much better setting until we change the way the number is chosen.

dk-teknologisk-mon commented 1 year ago

Aside from this lowering of suggested initial points, I also think it is a good idea to make it much clearer that you can change the number if you want to.