KI-Glossar
Das vollständige Wörterbuch der Künstlichen Intelligenz
CUSUM Algorithm
Sequential detection method based on the accumulation of residuals or cumulative sums to identify changes in the mean or variance of a process.
Page-Hinkley Detector
Optimal statistical test for detecting mean changes in a data stream, based on calculating the difference between the observed mean and the cumulative minimum mean.
Signal Segmentation
Process of partitioning a temporal signal into homogeneous segments where statistical properties remain constant between change points.
Abrupt Change
Instantaneous or quasi-instantaneous transition in the characteristics of a system, occurring on a very short time scale relative to the observation period.
Change Point
Precise instant in time when a structural break occurs, marking the transition between two distinct regimes of the studied system.
Online Detection
Real-time change detection, where the algorithm processes data sequentially as they arrive, without access to future observations.
Regime-Switching Model
Class of statistical models where parameters can switch between different states or regimes according to a stochastic or deterministic process.
Maximum Likelihood by Segment
Optimization technique consisting of separately maximizing the likelihood of each segment defined by potential change points.
Algorithme de Détecteur de Rupture
Ensemble de procédures computationnelles conçues pour identifier automatiquement les instants de changement dans les propriétés statistiques d'un processus.
Statistique de Changement
Variable aléatoire calculée à partir des observations et dont la distribution change significativement lorsqu'une rupture structurelle se produit.
Délai de Détection
Intervalle de temps entre l'occurrence réelle d'une rupture structurelle et sa détection effective par l'algorithme de surveillance.
Faux Positif en Détection
Identification incorrecte d'un point de rupture là où aucun changement structurel ne s'est produit, impactant négativement la robustesse du système de surveillance.