lkauto.optimization_strategies package

Submodules

lkauto.optimization_strategies.bayesian_optimization module

lkauto.optimization_strategies.bayesian_optimization.bayesian_optimization()

returns the best configuration found by bayesian optimization. The bayesian_optimization method will use SMAC3 to find the best performing configuration for the given train split. The ConfigurationSpace can consist of hyperparameters for a single algorithm or a combination of algorithms.

Parameters

  • train (pd.DataFrame) – Pandas Dataframe outer train split.

  • validation (pd.DataFrame) – Pandas Dataframe validation split.

  • cs (ConfigurationSpace) – ConfigurationSpace with all algorithms and hyperparameter ranges defined.

  • time_limit_in_sec (int) – time limit in seconds for the optimization process

  • num_evaluations (int) – number of samples to be drawn from the ConfigurationSpace

  • optimization_metric (function) – LensKit prediction accuracy metric to optimize for.

  • minimize_error_metric_val (bool) – Bool that decides if the error metric should be minimized or maximized.

  • user_feedback (str) – Defines if the dataset contains explicit or implicit feedback.

  • random_state (int) –

  • split_folds (int) – number of folds for cross validation

  • split_strategie (str) – cross validation strategie (user_based or item_based)

  • split_frac (float) – fraction of the dataset to be used for the validation split. If num_folds > 1, the fraction value will be ignored.

  • ensemble_size (int) – number of models to be used in the ensemble of rating prediction tasks. This value will be ignored for recommender tasks.

  • num_recommendations (int) – number of recommendations to be made for each user. This value will be ignored for rating prediction.

  • filer (Filer) – filer to manage LensKit-Auto output

Returns

  • incumbent (Configuration) – best configuration found by bayesian optimization

  • top_n_runs (pd.DataFrame) – top n runs found by bayesian optimization

Module contents