pystra.active_learning.surrogates.PCEFitResult

Use it: User guide · Worked example · Theory

class PCEFitResult(degree, q_norm, indices, coefficients, loo_error, corrected_loo_error, candidates, n_rank_deficient_bootstrap)

Bases: object

Immutable selected Hermite expansion and evaluated candidate scores.

indices contains one power tuple per coefficient in normal-coordinate variable order. Errors are normalized by population variance of training responses. LOO is conditional on the chosen support, not a complete cross-validation of model selection or a failure-probability error bound.

Methods