bonni.optimize_ipopt

Contents

bonni.optimize_ipopt#

bonni.optimize_ipopt(fn, x0, bounds, max_fn_eval=None, max_iterations=None, save_path=None, tol=1e-08, direction='minimize', bound_contract_ratio=0.001)[source]#

Optimize a black-box function using the Interior Point Optimizer (IPOPT).

This function acts as a convenient wrapper around cyipopt, managing the interface between the objective function (which must provide gradients) and the solver. It handles bound constraints, optimization direction, and recording of the optimization history.

Parameters:

  • fn (Callable[[np.ndarray], tuple[np.ndarray, np.ndarray]]) – The objective function to optimize. It must accept an input array of shape (num_dims,) and return a tuple (y, g), where y is the scalar objective value and g represents gradients of shape (num_dims,).

  • x0 (np.ndarray) – The initial starting point for the optimization. Must have shape (num_dims,).

  • bounds (np.ndarray) – A 2D array of shape (num_dims, 2) specifying the (lower, upper) search space boundaries for each of the num_dims input dimensions.

  • max_fn_eval (int | None, optional) – The maximum number of allowed function evaluations. Note that IPOPT performs line searches, so one iteration results in multiple function evaluations. Defaults to None.

  • max_iterations (int | None, optional) – The maximum number of solver iterations. Defaults to None.

  • save_path (Path | str | None, optional) – Directory path to save the results of function evaluation as an npz-file. Defaults to None.

  • tol (float, optional) – The tolerance for convergence. Defaults to 1e-8.

  • direction (Literal["maximize", "minimize"], optional) – The optimization goal. Defaults to “minimize”.

  • bound_contract_ratio (float, optional) – A small epsilon value used to contract the provided bounds slightly (bounds +/- eps * bound_range). This prevents the interior point method from starting or evaluating exactly on the boundary, which can cause numerical issues. Defaults to 1e-3.

Returns:

A tuple containing the full history of:

  • xs: Input parameters (shape (N, D)).

  • ys: Objective values (shape (N,)).

  • gs: Gradients (shape (N, D)).

Return type:

tuple[np.ndarray, np.ndarray, np.ndarray]

Contents