The issue #440 may have many causes, but one thing that is probably worth taking a note from there is that if uniform bounding is used and a bootstrap enlarge factor is large, that will lead to potentially very slow sampling. And I think we should put a warning in that case, as that's likely the case where ellipsoidal bounds are just not a good approximation to the L>L* volume.
I don't know the exact thresholds, but I'm suspecting that something like enlarge_factor^ndim > 10 would make sense.
I also don't know how likely those thresholds are to be breached.
segasai
commented
1 year ago
The issue #440 may have many causes, but one thing that is probably worth taking a note from there is that if uniform bounding is used and a bootstrap enlarge factor is large, that will lead to potentially very slow sampling. And I think we should put a warning in that case, as that's likely the case where ellipsoidal bounds are just not a good approximation to the L>L* volume.
I don't know the exact thresholds, but I'm suspecting that something like enlarge_factor^ndim > 10 would make sense. I also don't know how likely those thresholds are to be breached.