How to monitor convergence in classification and refinement?


I guess this will be covered in the forthcoming paper, but I’m curious as to what parameters one should pay attention to when assessing the progression of classification and refinement? And is there anyway to plot various parameters against the iteration number, to get an idea of how things are converging? Is there anyway in cryoSPARC to analyse how particles are changing assignments during classsification, or how divergent classes are from one another?


Hi Oli,

Sorry for the slow response on this. For refinement you should keep an eye on the estimated GSFSC resolution which is shown in the plots which are produced at each iteration. You can also look at the distributions of viewing directions to check for issues relating to preferential views.

In ab initio/classification at the moment there isn’t too much to watch with regards to convergence right now. We’d like to add some more diagnostics here eventually.