hookeai.material_model_finder.train.training.check_model_parameters_convergence

check_model_parameters_convergence(model_parameters_history_steps, convergence_tolerance=0.01, trigger_tolerance=5, min_hist_length=0, is_verbose=False)

Check convergence of model parameters.

Parameters:

• model_parameters_history_steps (dict) – History (item, list) of each model parameter (key, str).

• convergence_tolerance (float, default=0.01) – Minimum significative relative change.

• trigger_tolerance (int, default=5) – Number of consecutive steps without a significant relative change of all parameters to trigger convergence.

• min_hist_length (int, default=0) – Minimum history length required to trigger convergence criterion.

• is_verbose (bool, default=False) – If True, enable verbose output.

Returns:

• is_converged (bool) – If True, then all model parameters have converged, False otherwise.

• parameters_status (dict) – Convergence status data (item, dict) of each parameter (key, str). Convergence status data includes the current convergence status (‘is_converged’), the current absolute change (‘absolute_change’), and the current relative change (‘relative_change’). Unknown change values are set to None.