pystra.active_learning.stopping

Use it: User guide · Worked example · Theory

Stateless stopping policies evaluated against immutable fit histories.

Beta criteria follow Eqs. (2)-(3) of Moustapha, Marelli and Sudret (2022), doi:10.1016/j.strusafe.2021.102174. Prediction bands are surrogate diagnostics, not rigorous bounds on the true failure probability.

Classes

AllCriteria	Require every supplied policy's surrogate and sampling decisions.
BetaBounds	Require a small relative spread of surrogate-based beta endpoints.
BetaStability	Require small relative changes in beta across consecutive enrichments.
BootstrapBounds	Stop when the bootstrap Pf range / full-design Pf is small.
LearningThreshold	Require the selection threshold and an interior failure probability.
StoppingCriterion	Separate surrogate convergence from final sampling precision.