Minvalue must be positive in local refinement

I encounter this error when I try the new local refinement with the flooring parameters.

Iter000 is fine, then all the figure went to blank and raise this error.

Hi @zhenyu_tan,

Have you noticed this error on other similar runs, or does it only happen with these parameters and this data? You may want to re-run the job with the exact same parameters but “Reset input per-particle scale” activated, in case the input scales are incorrect. We’ve seen this error happen when there are numerical errors related to unexpected input data or parameter choices, or when there are very few particles.

For the parameter choices, I’ve generally seen that disabling the “Minimize over per-particle scale” parameter tends to give better results for masked refinements. As well, setting the search extents to be much larger than the prior standard deviations usually will result in the poses not changing very much at all. Typically the search extents should be on the order of 3 times that of the prior widths, so here, it would make sense to set the rotation search extent to 9 degrees and the shift search extent to 20 Angstroms. Could you let me know if the error persists after adjusting these parameters?


We have seen this issue, and the issue was (I think) that the mask had some small negative values (e.g -0.0001 or whatever) - I suspect some kind of integer/floating point conversion error in mask generation

Hi Ollie, hope you’re doing well! How did you get around this problem?

Hi Nick,

All good hope all is well with you too! In this case, just re-making the mask (re-thresholding and adding a soft edge) in cryosparc using Volume Tools did the trick.


1 Like

Cheers, that did the trick!

1 Like