Hi,
How are scale-rejected particles treated in refinement? Are their orientations still refined, or once they are rejected they are excluded from both refinement and reconstruction?
I ask because I have noticed that even with simulated data, where I know that all particles are “good”, a certain (small) percentage of particles are being rejected based on scale (presumably because they are being misaligned).
This makes me wonder if this might happen for some real data, particularly because scale refinement by default starts in the first iteration. If a particle is rejected in the first iteration, does it always remain rejected, or are the scales/orientations of even rejected particles re-evaluated at each iteration?
Cheers
Oli
2 Likes
Hey @olibclarke – all “used” scales are clipped at 0 during refinement so any negative scale particles are effectively invisible to the reconstruction, but their alignments are still updated.
This is re-evaluated at every iteration so particles can be “rejected” at a particular iteration and then later incorporated (we have seen this happen in our own testing). The rejected particles output is the latest set of rejected particles.
I ask because I have noticed that even with simulated data, where I know that all particles are “good”, a certain (small) percentage of particles are being rejected based on scale (presumably because they are being misaligned).
We have also seen this! I’ve noticed that even relatively (visually) minor misalignments in shifts can sometimes cause this.
2 Likes
Thanks Valentin, that is a very helpful explanation!