Number of epochs (Or batch size)


I wonder how many epochs are used in 3D Classification and Ab-Initio Reconstruction task.


Hi @minkyujeon

3D Class has an explicit parameter that sets the number of epochs used during the ‘online’ expectation maximization (O-EM) stage. Once these are complete, the job will also complete a dynamic number of ‘full batch’ EM iterations that will terminate based on convergence criteria.

Ab-initio doesn’t use the notion of ‘epochs’ explicitly; instead, the job will complete several hundreds of iterations with varying batch size. Both the number of iterations and the per-iteration batch size will depend on a number of parameters in the job. For many datasets, a single-class ab-initio job will use less than one ‘epoch’ of data (by design). In the multi-class case, ab-initio will always use at least one epoch of data.

Hope that helps!