pystra.active_learning.stopping.BootstrapBounds

Use it: User guide · Worked example · Theory

class BootstrapBounds(*, tolerance=0.1, consecutive=2, target_cov=0.1)

Bases: StoppingCriterion

Stop when the bootstrap Pf range / full-design Pf is small.

Marelli & Sudret (2018), Eqs. (8)-(9). Defaults: tolerance 0.1 and two consecutive fits. Uses actual replicate classifications on the fixed IID normal pool, not mean +/- two spread. This range is a bootstrap diagnostic, not a confidence interval accounting for model-selection bias. Final sampling precision is checked independently. Adaptive weighted/conditional enrichment does not yet supply bootstrap probability ranges and is rejected.

Methods

accepts_estimate	Require an interior probability and adequate finite sampling CoV.
should_stop	Apply the relative probability range over the trailing window.

should_stop(history)

Apply the relative probability range over the trailing window.

Return type:

bool

accepts_estimate(estimate)

Require an interior probability and adequate finite sampling CoV.

Return type:

bool