Training an Approximation with a Simulation Job

You can train an approximation using parameters from an existing simulation job, DOE adapter, or Monte Carlo adapter.

Before you begin: Run a simulation process with a DOE or Optimization.
See Also
About Approximation Training
Approximation Adapter
  1. From the Utilities section of the action bar, click Approximation Trainer and drop it on the process diagram.
  2. Double-click Approximation Trainer .
    The Approximation Trainer Editor opens.
  3. Click the Select Jobs tab and do either of the following to select the simulation job on which to train your approximation:

    • Select Content, and select the job on which to train your approximation.
    • Select Parameter, and choose an object parameter representing the simulation job or create a new simulation job object parameter.

    You can review the input and output parameters for the job in the table.

  4. Click the Configure Training tab.
  5. Enter a descriptive Approximation Object Name.
  6. Add a new configuration:

    Training an approximation requires the definition of at least one configuration. A configuration specifies an algorithm type, algorithm hyperparameters, validation type, and validation hyperparameters for a specified set of output parameters. Each output parameter can be included in only one configuration, and each configuration must include at least one output parameter.

    1. Click Add Configuration.
    2. From the Algorithm tab, configure the Approximation Type, along with any hyperparameters relevant to that type, such as a polynomial order for a Response Surface Model approximation.
    3. From the Validation tab, configure the Validation Type and hyperparameters.
    4. From the Outputs tab, select one or more output parameters to assign them to the configuration. If all parameters assigned to a configuration are reassigned to other configurations, the empty configuration is deleted.
    5. Click OK.
      The new configuration is created.
  7. Click OK.

The Approximation Trainer adapter creates an approximation object when it runs. You can use the approximation object to predict output parameters with the Approximation adapter.