tracts.driver.run_optimization#

run_optimization(physical_start_params, genetic_model, population, driver_spec, likelihood_options, model_param_names, sex_bias_param_names, print_run_details=True)#

Runs the optimization (single-step, or two-step with the step 1 / step 2 / step-2-skipped branches) given a list of starting parameters in physical units, returning the optimal parameters and the corresponding likelihood.

Parameters:

  • physical_start_params (list) – Starting parameter sets, in physical units, to start the optimization from (one optimization run per entry).

  • genetic_model (GeneticModel) – Bundles the demographic model being fit (accessible as genetic_model.demographic_model, which also provides parameter_handler for unit conversions and fixed-parameter state) with the admixture and phase-type model configuration used to compute the likelihood.

  • population (Population) – The population object containing the data to fit.

  • driver_spec (InferenceConfig) – The parsed driver-file configuration; controls whether a single-step or two-step optimization is run, and its iteration/verbosity settings.

  • likelihood_options (LikelihoodOptions) – Logging verbosity for this run. Its include_autosomes/include_allosomes flags are ignored here: which data are included is determined per step, as in optimize_cob_sex_biased_two_steps().

  • model_param_names (list) – Names of all of the demographic model’s free base parameters, in order.

  • sex_bias_param_names (list) – Names of the sex-bias parameters among model_param_names.

  • print_run_details (bool) – Whether to print the starting-parameters table under each step’s header, the “Optimization run #N” line, and the step-transition messages (“Selecting best parameters from step X…”). The step headers themselves and the optimizer’s own iteration table are always printed. Set to False for quick, low-noise repeated re-optimizations (e.g. run_sex_bias_fixing_reoptimizations) where only one run per step is performed and these details add clutter rather than information. Defaults to True.

Returns:

The optimal parameters found (in physical units) and the corresponding likelihood.

Return type:

tuple