adaptive.run_adaptive_search

run_adaptive_search(
    executor,
    input_cols,
    outcome_col,
    ranges,
    existing_data=None,
    n_start=10,
    n_step=5,
    max_iter=5,
    max_hours=None,
    max_gap_ratio=MAX_GAP_RATIO,
    min_r2_score=MIN_R2_SCORE,
    max_avg_cv=MAX_AVG_CV,
    max_max_cv=MAX_MAX_CV,
    output_csv=None,
    max_allowed_vif=MAX_ALLOWED_VIF,
    seed=None,
)

Orchestrates the Active Learning loop on raw DataFrames using the Executor architecture.

Parameters

Name	Type	Description	Default
executor	Executor | str	An instance of an Executor (Python, CLI, Matlab). Accepts a legacy command string for backward compatibility.	required
input_cols	List[str]	Input names.	required
outcome_col	str	Outcome name.	required
ranges	Dict	Input bounds.	required
existing_data	pd.DataFrame	Start data.	None
n_start	int	Init batch size.	10
n_step	int	Points added per refinement step.	5
max_iter	int	Max loops.	5
max_hours	float	Physical time limit in hours.	None
max_gap_ratio	float	Decimal Percentage threshold for maximum gap diagnostics,	MAX_GAP_RATIO
min_r2_score	float	Decimal Percentage threshold for minimum r2 diagnostics,	MIN_R2_SCORE
max_avg_cv	float	Decimal Percentage threshold for average CI diagnostics,	MAX_AVG_CV
max_max_cv	float	Decimal Percentage threshold for maximum CI diagnostics	MAX_MAX_CV
output_csv	str	Path to write the output CSV results incrementally.	None
max_allowed_vif	float	Collinearity (VIF) threshold. A collinearity failure cannot be fixed by sampling, so it stops the loop with a warning.	MAX_ALLOWED_VIF
seed	int	Seed for the initial design and every refinement step, for reproducible runs.	None

Returns

Name	Type	Description
	pd.DataFrame	pd.DataFrame: Final dataset containing all successful runs.