paulknysh/blackbox

Feature Suggestion: Load Past Results

Opened this issue · 2 comments

So I have a really long running function, and as a result, I am exploring a broad range of parameter settings (n is large, m is 0). But if it turns out I am not happy with the exploratory performance of this sweep, then I would like to go back, use those results, and run again to explore points near the top performers (n = 0, m is large).

Thanks, that might be a good idea!

I'm currently looking for adding this feature.
I'm tried to read through the paper but can't understand everything so I'm facing a problem. When I run the procedure until the subsequent iterations started and save the state, are there any complications when starting at the initial sampling again? Like a possible distortion of the result?