gaussian_ar

Gaussian autoregressive estimating schema.

class pysatl_cpd.algorithms.online.cusum.component.estimator.gaussian_ar.EstimatesGaussianAR[source]

Bases: ISchemaEstimates

intercept: float

coefficients: UnivariateNumericArray

noise_variance: float

history: UnivariateNumericArray

class pysatl_cpd.algorithms.online.cusum.component.estimator.gaussian_ar.GaussianARSchema(autoreg_order=1, adaptive=True, window=None)[source]

Bases: IEstimatingSchema[UnivariateNumericArray, EstimatesGaussianAR]

Univariate autoregressive estimation schema backed by arch ARX.

Parameters:

autoreg_order (int) – Number of lags in the AR model (> 0).
adaptive (bool) – Whether to re-fit the model online after training.
window (int | None) – Maximum number of observations retained for fitting. None means unbounded.

Raises:

ValueError – If autoreg_order is non-positive, or window is non-positive.
RuntimeError – If the optional arch dependency is not installed.

__init__(autoreg_order=1, adaptive=True, window=None)[source]

Parameters:

autoreg_order (int)
adaptive (bool)
window (int | None)

Return type:

None

train(train_set)[source]

Fit AR model parameters from a training sample.

Parameters:: train_set (Sequence[ndarray[tuple[int], dtype[double]]]) – Training observations.
Return type:: None

update(observation)[source]

Update the internal data buffer and optionally re-fit AR weights.

Parameters:: observation (ndarray[tuple[int], dtype[double]]) – New observation.
Return type:: None

property estimates: EstimatesGaussianAR

Current AR model estimates.

Return type:: EstimatesGaussianAR

reset()[source]

Reset all AR parameters and data buffer.

Return type:: None