gaussian

Gaussian monitoring schema for generalized CUSUM.

This module provides GaussianMonitoringSchema, which whitens observation residuals using the inverse square root of covariance.

class pysatl_cpd.algorithms.online.cusum.component.monitoring.gaussian.GaussianMonitoringSchema(cov_reg)[source]

Gaussian monitoring transformation based on mean and covariance estimates.

Parameters:: cov_reg (float) – Diagonal covariance regularization added before matrix inversion.

__init__(cov_reg)[source]

evaluate(observation, parameters)[source]

Transform observation into whitened monitoring residual.

Computes cov^{-1/2} @ (observation - mean).

Parameters:

observation (ndarray[tuple[int], dtype[double]]) – New observation vector.
parameters (EstimatesGaussianMLE) – Estimates dict containing "mean" and "cov".

Returns:

Whitened residual vector.

Return type:

ndarray[tuple[int], dtype[double]]

Reset internal dimensionality tracker.

Notes

The monitoring transform itself is stateless; only the cached dimension is reset.