When using this framework with external evaluation code (that uses network or scripts written in another languages), it is crucial and very important to have ability to evaluate candidates in batches.
One of the examples of such need is when you're calling external script that have large "heating up" time (due to library loading, etc) that is larger than useful evaluation work that it does.
Another example is when your generational population data stored in network database and worker processes that evolve it are residing on different computers — so it's too expensive (time-wise) to transfer same populace-bound data for evolution of each candidate.
When using this framework with external evaluation code (that uses network or scripts written in another languages), it is crucial and very important to have ability to evaluate candidates in batches.
One of the examples of such need is when you're calling external script that have large "heating up" time (due to library loading, etc) that is larger than useful evaluation work that it does.
Another example is when your generational population data stored in network database and worker processes that evolve it are residing on different computers — so it's too expensive (time-wise) to transfer same populace-bound data for evolution of each candidate.