optimize_category_assignment_parallel#

obsidian.experiment.advanced_design.optimize_category_assignment_parallel(cat_samples, conditional_subparameters, subparam_mapping, n_samples, seed, max_workers=4, n_category_trials=100, corr_threshold=0.01)[source]#

Search for the category assignment (for the variable that has a subparam mapping) that minimizes the maximum inter-category correlation.

Parameters:
  • cat_samples – Dict of already-sampled categorical columns (excluding the variable being optimized).

  • conditional_subparameters – Full conditional subparameter specification.

  • subparam_mapping – Dict mapping each categorical variable to its subparameter.

  • n_samples – Number of rows in the design.

  • seed – Base random seed.

  • max_workers – Maximum number of parallel threads. Defaults to 4.

  • n_category_trials – Number of random assignments to evaluate. Defaults to 100.

  • corr_threshold – Stop early if max correlation drops below this value. Defaults to 0.01.

Returns:

Best-found list of category values for the optimized variable.

Return type:

list