baybe.simulation.scenarios.simulate_scenarios

baybe.simulation.scenarios.simulate_scenarios(scenarios: dict[Any, Campaign], lookup: DataFrame | Callable[[DataFrame], DataFrame] | None = None, /, *, batch_size: int = 1, n_doe_iterations: int | None = None, initial_data: list[DataFrame] | None = None, groupby: list[str] | None = None, n_mc_iterations: int | None = 1, random_seed: int | None = None, impute_mode: Literal['error', 'worst', 'best', 'mean', 'random', 'ignore'] = 'error', noise_percent: float | None = None, parallel_runs: bool | None = None)

Simulate multiple Bayesian optimization scenarios.

A wrapper function around baybe.simulation.core.simulate_experiment() that allows to specify multiple simulation settings at once.

Parameters:

• scenarios (dict[Any, Campaign]) – A dictionary mapping scenario identifiers to DOE specifications.

• lookup (Union[DataFrame, Callable[[DataFrame], DataFrame], None]) – See baybe.simulation.core.simulate_experiment().

• batch_size (int) – See baybe.simulation.core.simulate_experiment().

• n_doe_iterations (Optional[int]) – See baybe.simulation.core.simulate_experiment().

• initial_data (Optional[list[DataFrame]]) – A list of initial data sets for which the scenarios should be simulated.

• groupby (Optional[list[str]]) – The names of the parameters to be used to partition the search space. A separate simulation will be conducted for each partition, with the search restricted to that partition.

• n_mc_iterations (int | None) – The number of Monte Carlo simulations to be used. If set to None, one Monte Carlo simulation per initial data set is conducted, but the random seed is incremented with each initial data set.

• random_seed (Optional[int]) – An optional integer specifying the random seed for the first Monte Carlo run. Each subsequent runs will increase this value by 1. If omitted, the current random seed is used.

• impute_mode (Literal['error', 'worst', 'best', 'mean', 'random', 'ignore']) – See baybe.simulation.core.simulate_experiment().

• noise_percent (Optional[float]) – See baybe.simulation.core.simulate_experiment().

• parallel_runs (Optional[bool]) – If set, the value is handed over to the parallel argument of combo_runner(), enabling the use of parallel runners. When omitted, its value is determined by the BAYBE_PARALLEL_SIMULATION_RUNS environment variable, with fallback to True.

Return type:

DataFrame

Returns:

A dataframe like returned from baybe.simulation.core.simulate_experiment() but with additional columns. See the Note for details.

The following additional columns are contained in the dataframe returned by this function:

• Scenario: Specifies the scenario identifier of the respective simulation.

• A column Random_Seed that specifies the random seed used for the respective simulation.

• A column Initial_Data that specifies the index of the initial data set used for the respective simulation or contains NaN if no initial data is provided.

• Optional, if groupby is provided: A column for each groupby parameter that specifies the search space partition considered for the respective simulation.