Hi,
In heterogeneous refinement, convergence statistics (class switching %s) are printed to the log for O-EM iterations, but not for full iterations. It might be useful if this was also printed for full iterations, so the user can judge whether additional full iterations might be useful/desirable.
This isn’t normally the case, but can be for particularly large datasets, especially of small/low SNR particles.
(The same applies for 2D classification actually - sometimes particularly small particles benefit from many full iterations - convergence stats might be helpful there too.)
Cheers
Oli