tracts.driver_utils.build_boundary_reoptimization_model#

build_boundary_reoptimization_model(driver_spec, reload_context, boundary_fixed_param_values, genetic_model, optimal_params, remainder_params, alternate_implicit_population=None)#

Builds a genetic model and driver specification identical to the given ones, except with the sex-bias parameters in boundary_fixed_param_values now fixed by value, and (if given) alternate_implicit_population as the implicit population instead of the current one. Fixed parameters and starting parameters are set up accordingly. Used to retry the optimization when one or more sex-bias parameters have an optimal value at a +-1 boundary.

As with _get_driver_for_reoptimization, all starting parameters are pinned to their previous optimal value (no resampling) and repetitions are set to 1: this re-optimization should resume from where the previous one left off, not restart from a fresh random point.

Parameters:

  • driver_spec (InferenceConfig) – The original driver-file configuration.

  • reload_context (ModelReloadContext) – File-location and ancestry-proportion context needed to reload the demographic model from the model YAML file, used only when alternate_implicit_population is given.

  • boundary_fixed_param_values (dict[str, float]) – Directly-optimized sex-bias parameter names (and their boundary value) to fix by value, merged into driver_spec.optim.fix_parameters_by_value.

  • genetic_model (GeneticModel) – The current genetic model. When alternate_implicit_population is None, a deep copy of this genetic model is reused instead of reloading the demographic model from the model YAML file (its parameter names are derived directly from genetic_model.demographic_model, which is unchanged in that case). When alternate_implicit_population is given, only its phase_type_config is reused (the demographic model must be reloaded, since which population is implicit is baked into the founder events at YAML-parse time).

  • optimal_params (ndarray) – The current optimal parameters (physical units), before this re-optimization, in the order given by genetic_model.demographic_model’s free base parameters. Used to pin the starting parameters of the re-optimization (see _get_driver_for_reoptimization).

  • remainder_params (dict[str, float]) – The remainder (derived) parameters computed from the previous optimization’s optimal parameters, as returned by compute_remainder_params. Used, when alternate_implicit_population is given, to set a starting value for the rate, and fix by value the sex-bias parameter, of the population that was previously implicit and is now explicit (see _get_params_for_newly_explicit_population).

  • alternate_implicit_population (str | None) – The name of the population to use as the implicit population instead of the current one. If None, the current implicit population is kept. Defaults to None.

Returns:

The new driver spec; the new genetic model built from it; the model, sex-bias, and non-sex-bias parameter names; and the physical starting parameters to optimize from.

Return type:

tuple[InferenceConfig, GeneticModel, list[str], list[str], list[str], list[np.ndarray]]