Hi,
In 2D and 3D classification, it is sometimes useful to limit the minimum resolution for alignment - kind of like a highpass filter.
This has been implemented for 2D and 3D classification, and we find it useful in certain cases (small particles, subtle features).
It seems intuitively that this might also be useful for ab initio reconstruction of small particles, where low resolution features may be ambiguous or unreliable - would it be possible to implement there too? It is currently possible to set the initial and final maximum resolution, but no minimum resolution cutoff.
I have tried some preliminary tests with Aca2-RNA, and in that case, ab initio certainly converges faster with the same parameters and random seed when using a 100Å highpass filtered stack (applied using the Downsample job) than using the original stack of the same size.
Cheers
Oli