3D classification: query regarding PCA/sample particles per reconstruction parameter

Hi,

Concerning the setting “PCA/simple particles per reconstruction”, the description states that ranges of 1000-5000 are best for “simple” mode, or 100-500 for PCA. Is there some reason to use less particles for density initialization with PCA? Or is this purely for optimizing speed?

Cheers
Oli

Hey @olibclarke,

Nope, you are right on – this recommendation is largely there for computational efficiency reasons. In internal testing, we haven’t seen a significant difference in PCA mode initializations for larger settings of ‘particles per reconstruction’. Since computing 100+ reconstructions with 1000+ particles each can be time consuming, I kept the recommendation in the current form. Please do let me know if you find it useful to use larger values for different datasets.

Valentin

1 Like

Great, thanks for confirming Valentin!

Cheers
Oli