Dear ccpem list,
I have noticed during the first two iterations of either classification or refinement runs that the map SSNR was sometimes calculated only up to the starting resolution, hence no improvement in resolution was observed for those iterations, and sometimes it was calculated on the full achivable resolution for the particular box size at that iteration, hence resolution improved. The behavior seem to be random as reapeated runs with the same dataset could give different results.
It can be an issue, as the resolution often improves dramatically at the subsequent iteration, with the expectation time going from minutes to hours.
Any light on this would be appreciated.