Difference between input and output particle #


I ran an ab initio job with ~220k particles and asked for 3 classes. One class, which accounted for ~50% of the particles, was very nice and I used this for further experiments.

However, in subsequent jobs, this class only contains 70k particles, instead of an expected 110k. If I total the number of particles in the 3 classes, it only accounts for 130/220k particles

Any insight into what is happening? I thought every particle would be put into a class. Are particles being excluded?


(If you look at the three classes in the attached picture, you see that they only contain 133,327 out of an expected 218,420)

This is due to the class probability threshold - if you click select “input classes”, you’ll see there is a class probability threshold which by default is 0.9. Which means that of all the particles that are nominally in that class, only those with an assignment probability of 0.9 or greater will be included in the subsequent refinement (this is actually a useful parameter to tune for getting a cleaner set of particles - increasing it to 0.99 sometimes gives better results).


Awesome, thanks. I’ll play around with that parameter.