pystra.plotting.plot_surrogate_slice#

plot_surrogate_slice(surrogate, points, *, coordinate=0, limit_state=None, n_std=2.0, batch_size=2048, ax=None)[source]#

Plot a fitted surrogate along one independent-normal coordinate.

Parameters:

  • surrogate (Surrogate) – Fitted surrogate; prediction only, without training or RNG draws.

  • points (ndarray) – Slice points. The selected coordinate must strictly increase; all other coordinates must remain fixed. Rows are independent-normal coordinates.

  • coordinate (int) – Varying coordinate column.

  • limit_state (Callable | None) – True response with batch input (n, n_variables) and output (n,). Must use the same coordinates; these n evaluations are additional cost.

  • n_std (float) – Surrogate-spread multiplier, not a confidence level. Bootstrap spread and Kriging predictive std have different meanings; neither establishes true-model coverage or includes all model-selection uncertainty.

  • batch_size (int) – Maximum rows in each prediction/evaluation call.

  • ax (Axes | None) – Existing axes.

Returns:

figure, ax – Figure and axes, without displaying them.

Return type:

tuple[Figure, Axes]