Class PELTPenaltySelector

Namespace

SignalSharp.Detection.PELT

Assembly

SignalSharp.dll

Provides methods to automatically select a penalty for the PELT algorithm using various criteria like BIC, AIC, or AICc.

public class PELTPenaltySelector

Inheritance

object

PELTPenaltySelector

Inherited Members

object.Equals(object)

object.Equals(object, object)

object.GetHashCode()

object.GetType()

object.MemberwiseClone()

object.ReferenceEquals(object, object)

object.ToString()

Remarks

Penalty selection is crucial for controlling the number of change points detected by PELT. This class implements common information criteria for model selection based on likelihood principles.

Note that likelihood-based criteria (BIC, AIC, AICc) require the cost function used in the PELT algorithm to implement ILikelihoodCostFunction correctly and set SupportsInformationCriteria to true. The PELT algorithm instance provided to this selector must be fitted with the signal data *before* calling the `SelectPenaltyInternal` method (or implicitly via `FitAndSelect`).

Constructors

PELTPenaltySelector(IPELTAlgorithm)

Initializes a new instance of the PELTPenaltySelector class.

public PELTPenaltySelector(IPELTAlgorithm peltAlgorithm)

Parameters

peltAlgorithm IPELTAlgorithm

The PELT algorithm instance to use. Its internal state (e.g., fitted cost function) will be used.

Exceptions

ArgumentNullException

Thrown if peltAlgorithm, its Options, or its CostFunction is null.

ArgumentOutOfRangeException

Thrown if MinSize or Jump in peltAlgorithm.Options are invalid.

Properties

PELTAlgorithm

Gets the PELT algorithm instance used for penalty selection.

public IPELTAlgorithm PELTAlgorithm { get; }

Property Value

IPELTAlgorithm

Methods

FitAndSelect(double[,], PELTPenaltySelectionOptions)

Fits the internal PELT algorithm to the multi-dimensional signal matrix and selects the optimal penalty based on the specified options.

public PELTPenaltySelectionResult FitAndSelect(double[,] signalMatrix, PELTPenaltySelectionOptions selectionOptions)

Parameters

signalMatrix double[,]

The multi-dimensional signal data (rows=dimensions, cols=time) used to fit the PELT algorithm.

selectionOptions PELTPenaltySelectionOptions

Configuration for the penalty selection process (method, range, etc.).

Returns

PELTPenaltySelectionResult

A PELTPenaltySelectionResult containing the selected penalty and corresponding breakpoints.

Exceptions

ArgumentNullException

Thrown if signalMatrix or selectionOptions are null.

InvalidOperationException

Thrown if a likelihood-based selection method is requested but the cost function does not support it.

PELTAlgorithmException

If an error occurs during the selection process or if no suitable penalty is found.

FitAndSelect(double[], PELTPenaltySelectionOptions)

Fits the internal PELT algorithm to the 1D signal and selects the optimal penalty based on the specified options.

public PELTPenaltySelectionResult FitAndSelect(double[] signal, PELTPenaltySelectionOptions selectionOptions)

Parameters

signal double[]

The 1D signal data used to fit the PELT algorithm.

selectionOptions PELTPenaltySelectionOptions

Configuration for the penalty selection process (method, range, etc.).

Returns

PELTPenaltySelectionResult

A PELTPenaltySelectionResult containing the selected penalty and corresponding breakpoints.

Exceptions

ArgumentNullException

Thrown if signal or selectionOptions are null.

InvalidOperationException

Thrown if a likelihood-based selection method is requested but the cost function does not support it.

PELTAlgorithmException

If an error occurs during the selection process or if no suitable penalty is found.