Dakota Reference Manual  Version 6.13
Explore and Predict with Confidence
 All Pages
batch_size


The number of evaluation points to include in each batch.

Specification

Alias: none

Argument(s): INTEGER

Child Keywords:

Required/Optional Description of Group Dakota Keyword Dakota Keyword Description
Optional exploration

Portion of batch size dedicated to exploration

Description

The number of evaluation points to include in each batch, typically specified for purposes of concurrent execution of these points.

Examples

In the context of adaptive surrogate refinement, batch size requests the generation of multiple refinement candidates that can be evaluated concurrently. This may represent a trade-off between algorithmic efficiency (selecting only the best refinement candidate, one at a time) and scalability (selecting multiple refinement candidates, of decreasing value, to maintain a desired concurrency/load balance).