Complete concurrent evaluation of all evaluations within a batch


  • Alias: None

  • Arguments: None


In the blocking case, all of the concurrent evaluations in the batch must be completed and all results are returned together to the algorithm.

These runs are reproducible, assuming use of the same seed in cases with random number generation.


  • In the specific context of parallel pattern search, if the best of these trial points is an improving point, then it becomes the next iterate.

  • In the specific context of adaptive refinement of Gaussian process (GP) models, all look-aheads are replaced with truth data and a new set of refinement candidates can then be generated based on the updated state of the GP.