SharpBCI.TournamentArtifactDetector Class Reference

The TournamentArtifactDetector holds a "tournament" of some number ARArtifactDetectors (termed competitors) The majority consensus of the most well-fit models is the final determination of whether or not the data is an artifact. More...

Inheritance diagram for SharpBCI.TournamentArtifactDetector:

Public Member Functions
	TournamentArtifactDetector (uint tournamentSize, uint learningSetSize, uint nAccept, uint initialMerits)
double	Error ()
	Inform the caller of the current confusion of this implementation The returned value should be a double greater than or equal to zero Otherwise no range is assumed, but lower values should indicate lower confusion For example, a detector that perfectly models the signal with no artifacts should have an error of zero Pragmatically, error will generally be non-zero due to presence of artifacts and noise. More...
bool	Detect (double data)
	Update artifact detector using next. More...

Protected Member Functions
IArtifactDetector	NewCompetitor ()

Detailed Description

The TournamentArtifactDetector holds a "tournament" of some number ARArtifactDetectors (termed competitors) The majority consensus of the most well-fit models is the final determination of whether or not the data is an artifact.

The detector defines a concept called "(de)merits". (De)merits are points which are added when the competitor's error is low and taken away when the competitor's error is high. If the (de)merits of a competitor is less than zero, it will be ejected from the tournament and replaced with a newly fit competitor. Thus, every once in a while, the artifact detector will potentially re-fit some number of competitors. The time interval, in the worst case, between re-fits of the model is determined by initialMerits. In this way, the TournamentArtifactDetector will converge on the best fit for the underlying process while this process is wide-sense stationary. If the underlying process changes model, the TournamentArtifactDetector will eventually converge on the best-fit for the new process.

Constructor & Destructor Documentation

SharpBCI.TournamentArtifactDetector.TournamentArtifactDetector	(	uint	tournamentSize,
		uint	learningSetSize,
		uint	nAccept,
		uint	initialMerits
	)

Parameters

tournamentSize	the number of competitors in the tournament
learningSetSize	the size of the dataset used to fit new competitors
nAccept	how many competitors should be used to determine the consensus opinion of the detector
initialMerits	competitors which have high error will be ejected

Member Function Documentation

bool SharpBCI.TournamentArtifactDetector.Detect

(

double

next

)

Update artifact detector using next.

Returns

true iff next is an artifact, false otherwise

Implements SharpBCI.IArtifactDetector.

double SharpBCI.TournamentArtifactDetector.Error

(

)

Inform the caller of the current confusion of this implementation The returned value should be a double greater than or equal to zero Otherwise no range is assumed, but lower values should indicate lower confusion For example, a detector that perfectly models the signal with no artifacts should have an error of zero Pragmatically, error will generally be non-zero due to presence of artifacts and noise.

Returns

the internal confusion metric of this implementation

Implements SharpBCI.IArtifactDetector.

IArtifactDetector SharpBCI.TournamentArtifactDetector.NewCompetitor ( )

protected

---

The documentation for this class was generated from the following file:

• /home/travis/build/berkona/SharpBCI/SharpBCI/Artifacts.cs