Particle stack order and randomisation


I wanted to ask how the particle stacks are currently ordered, for example after 2D or 3D and in what order are the particles loaded into the subsequent job:

  1. For example, I run 2D classification and I get 3 good classes with 200k particles, 50k particles, and 5k particles in them, respectively. If I wanted to run a 2D classification afterwards with these selected classes, would the class with 200k particles be classified first?

  2. If yes, would this bias the classification in any way?

  3. How are the particles ordered after 3D refinement and could the order affect downstream jobs like ab-initio or heterogeneous refinement?

  4. Is there a possibility to randomise the stack?

Thank you very much!

Pretty sure a random, non-overlapping subset is chosen at each iteration, at least for 2D (in the first n-1 iterations)


@mpacesa Oli is right, the particles are randomized in each job where they’re used. Internally, the particles datasets are always stored in the order they were found during a picking step (even after going through 2D Classification or multi-class Ab-initio.

If you do want to randomize the the particle dataset, you may do so with the “Particle Sets” job: Set a split of 1, the batch size equal to the size of the particle dataset, and enable the randomize flag. Bug again, this is not required.

1 Like