pystra.active_learning.results.ReliabilityEstimate

Use it: User guide · Worked example · Theory

class ReliabilityEstimate(failure_probability, sampling_cov, sampling_interval, confidence_level, n_samples, method, sampling_dependence, converged=True, status='completed', diagnostics=())

Bases: object

Conditional failure estimate returned by a reliability estimator.

sampling_cov quantifies sampling error conditional on the fitted surrogate; it excludes surrogate bias. Infinity denotes unresolved precision. sampling_interval is either None or a (lower, upper) pair at confidence_level. Its calculation belongs to the estimator, as do importance weights and corrections for dependent samples. method names that calculation; sampling_dependence is ‘independent’ or ‘dependent’. n_samples counts surrogate classifications, not true limit-state evaluations.

converged means the estimator reached its target event, separately from satisfying a requested CoV. status explains incomplete sampling. diagnostics retains immutable method-specific records (SubsetRun for replicated subset simulation). Incomplete estimates must not be accepted as successful reliability results.

Methods

property beta: float

Normal-equivalent index; infinite at probability zero or one.