iterative ref: branch and bound

[geoffwoollard]

• way to slice up SO(3) (rotations) to do grid search
• input (grid search), output 2D rotated planes

[geoffwoollard]

See discussion in

• Scheres, S. H. W. (2012). RELION: Implementation of a Bayesian approach to cryo-EM structure determination. Journal of Structural Biology, 180(3), 519–530. http://doi.org/10.1016/j.jsb.2012.09.006
• Punjani, A., & Fleet, D. (2021). 3D Flexible Refinement : Structure and Motion of Flexible Proteins from Cryo-EM. BioRxiv, 1–21. http://doi.org/10.1101/2021.04.22.440893

[geoffwoollard]

When implementing it would be useful to benchmark what are the bottlenecks

• making rotations
• rotating xy_0 plane by rotations
• interpolating slices
• computing bayes factors of a particle on slices

Depending on how things are parallelized, and if the ctf is included or not in the bayes factor computations, some things could be reused between particles like the (library of slices without the ctf applied.

How to decide how finely gridded to make the grid search? By what criteria? Perhaps heuristics based on how well the signal agrees in half_map_1, half_map_2 (Fourier voxels).

[geoffwoollard]

See comments on healpy and benchmarking in #13