In the Simulate Data (BETA) job, cryoSPARC sample each of three angles independently from Gaussian dist (when number of gridpoint >= 4). However the probability of some combinations become so small (such as 0.010.010.01=1e-6) that can’t be sampled. It would be nice to sample projection from a uniform distribution over SO(3).
By the way, the job would fail when the generate file gets larger than ~15GB, how about split into multiple files?
As I’m sure many are aware, you can easily get a uniform distribution over SO(3) from normally distributed 4-vectors.
q = np.random.randn(N, 4)
q /= np.linalg.norm(q, axis=1)
but this case is one where a fixed set of angles can really come in handy, e.g. HealPix.
Later I find that increasing gridpoint to 160 results in a continuous distribution and the reconstructed map is comparable to experimental data.
Could you tell me the detail about the paras of Simulate data Job 
, I cant find the description about the paras, such as the gridpoint, dose it mean the same as the projection angle? and what dose “rotation between 2 bodies” mean ?
Thanks a lot !
Could you tell me the detail about the paras of Simulate data Job , I cant find the description about the paras, such as the gridpoint, dose it mean the same as the projection angle? and what dose “rotation between 2 bodies” mean ?
Thanks a lot !