KSLCore/ksl.utilities.distributions.fitting.mixture.refine

Package-level declarations

Types

class BreakShiftRefiner(val maxShift: Int = defaultMaxShift, val maxIterations: Int = defaultMaxIterations, val criterion: MixtureCriterionIfc = MixtureBICCriterion(), val selector: FamilySelectorIfc = SeparableCMLSelector()) : PartitionRefinerIfc

Improves a partition by moving one cut at a time to a nearby position, keeping any move that improves the criterion.

class ClassificationEMRefiner(val tolerance: Double = defaultTolerance, val maxIterations: Int = defaultMaxIterations) : PartitionRefinerIfc

Improves a partition by alternating exact reassignment with refitting, in the manner of classification expectation maximization.

Returns the initial partition unchanged.

Improves an initial partition with respect to a fitting objective.

Why a refinement stopped.

data class RefinementResult(val partition: DataPartition, val outcome: RefinementOutcome, val numIterations: Int, val objectiveTrace: List<Double>)

The result of refining a partition.