FIND THE OPTIMUM DESIGN BASED ON OBJECTIVE

Applies simulated annealing to find the optimal design point within a dataset by fitting a surrogate model (polynomial regression, Kriging, or curve interpolation) over the input/target columns and iteratively searching for the minimum or maximum objective. Use this worker when you need a stochastic, gradient-free optimizer to escape local minima across a design space defined by tabular data.

When to use

Classification: process.

Tagged: curve_interpolate, design_exploration, design_optimization, kriging, objective_function, optimization, polynomial_regression, simulated_annealing.

Inputs

Label ID Type Default Required Description
Dataset dataset_1 dataset   Input dataset (tabular) containing design variable columns and response/target columns; all subsequent column selections must refer to column names present in this dataset.
Inputs inputs text   One or more dataset column names to treat as design input variables (X) fed into the surrogate model; select all independent variable columns.
Targets targets text   One or more dataset column names to treat as response/target variables (Y) that the optimizer will minimize or maximize; select the objective-carrying columns.
Number Of Iterations num_iterations scalar 100   Total number of simulated-annealing iterations to run; default is 100 — increase for higher-dimensional or noisy spaces at the cost of compute time.
Step Size step_size scalar 0.01   Perturbation magnitude applied to input variables at each iteration (dimensionless fraction of the normalized input range); default 0.01 — reduce for fine local search, increase for broader exploration.
Initial Temperature initial_temperature scalar 10   Starting temperature for the annealing schedule controlling the probability of accepting worse solutions early in the search; default 10 — higher values increase early exploration.
Model Type model_type scalar polynomialRegression   Surrogate model used to evaluate the objective between data points; choose ‘polynomialRegression’, ‘kriging’, or ‘curve_interpolate’ — default is ‘polynomialRegression’.
Model Parameters model_params form-input-table    
Shrink Factor shrink_factor scalar 0    
Return Type return_type scalar optimum    
Data Normalization Type normalize scalar minmax    
Target Value target_value textarea 0   Target value for optimum
Objective objective scalar minimize    
Grid Search grid_search scalar no    
Constraints constraints dataset   Dataset with constraints. The expected columns are needle, condition, target. Where needle is the column that is meet the condition specified by target. Ex needle=variable1 condition=gt and target=10.0 <a class=’btn btn-xs btn-default’ target=’_blank’ href=’https://www.d3view.com/docs/master/workflows/Glossary.html#datasetinput’> <i class=’fa fa-external-link’> </i> View more </a>
Mathmodel mathmodel_id remote_lookup    
STOP RATIO stopping_criteria scalar 0.1   When convergence is not improving after STOP_RATIO*NUM_ITERATIONS, the optimization is terminated. If STOP < 0, the optimization is stopped when convergence is not improving within abs(STOP_RATIO) steps

Outputs

Label ID Type Description
dataset_simulated_annealing_optimizer_output_1 dataset_simulated_annealing_optimizer_output_1 dataset Output dataset containing the optimum design point(s) found by the annealing search, including the optimal input variable values and the corresponding predicted response value(s).

Disciplines

  • ai_ml.surrogate
  • data.dataset.transform
  • design_exploration.optimization

Runnable example

A runnable example is registered for this worker. Open the example workflow on the d3VIEW canvas: /api/workflow/example?id=dataset_simulated_annealing_optimizer


Auto-generated from transformation schema. Worker id: dataset_simulated_annealing_optimizer. Schema hash: 7a9f3205e2f3. Hand-curated docs in workerexamples/ override this page when present.