multidim_parameter_study

Samples variables on full factorial grid of study points

Topics

parameter_studies

Specification

  • Alias: None

  • Arguments: None

Child Keywords:

Required/Optional

Description of Group

Dakota Keyword

Dakota Keyword Description

Required

partitions

Samples variables on full factorial grid of study points

Optional

model_pointer

Identifier for model block to be used by a method

Description

Dakota’s multidimensional parameter study computes response data sets for an n-dimensional grid of points. Each continuous and discrete range variable is partitioned into equally spaced intervals between its upper and lower bounds, each discrete set variable is partitioned into equally spaced index intervals. The partition boundaries in n-dimensional space define a grid of points, and every point is evaluated.

Default Behavior

By default, the multidimensional parameter study operates over all types of variables.

Expected Outputs

A multidimensional parameter study produces a set of responses for each parameter set that is generated.

Expected HDF5 Output

If Dakota was built with HDF5 support and run with the hdf5 keyword, this method writes the following results to HDF5:

Usage Tips

Since the initial values from the variables specification will not be used, they need not be specified.

Examples

This example is taken from the Users Manual and is a good comparison to the examples on centered_parameter_study and vector_parameter_study.

# tested on Dakota 6.0 on 140501
environment
  tabular_data
    tabular_data_file = 'rosen_multidim.dat'

method
  multidim_parameter_study
    partitions = 10 8

model
  single

variables
  continuous_design = 2
    lower_bounds     -2.0     -2.0
    upper_bounds      2.0      2.0
    descriptors       'x1'     "x2"

interface
  analysis_driver = 'rosenbrock'
    fork

responses
  response_functions = 1
  no_gradients
  no_hessians

This example illustrates the full factorial combinations of parameter values created by the multidim_parameter_study. With 10 and 8 partitions, there are actually 11 and 9 values for each variable. This means that \(11 \times 9 = 99\) function evaluations will be required.