pystra.active_learning.surrogates.KrigingSurrogate

Use it: User guide · Worked example · Theory

class KrigingSurrogate(*, n_restarts=2, noise=1e-10, seed=None)

Bases: Surrogate

Matérn 5/2 Gaussian process with normalized response.

Parameters are keyword-only. noise is a positive numerical nugget; n_restarts controls additional likelihood optimization starts. seed owns the scikit-learn optimizer’s random state. Requires pip install 'pystra[al]'. Inputs are normal coordinates.

Methods

fit	Fit normal-coordinate observations; reject nonfinite inputs.
predict	Return posterior mean/std; raise RuntimeError before fitting.

fit(points, values)

Fit normal-coordinate observations; reject nonfinite inputs.

Return type:

None

predict(points)

Return posterior mean/std; raise RuntimeError before fitting.

Return type:

tuple