dk.bayes.factorgraph.ep.calibrate

fb

package fb

Visibility
  1. Public
  2. All

Type Members

  1. case class EPSummary(iterNum: Int, msgNum: Long) extends Product with Serializable

    Expectation Propagation execution summary.

    Expectation Propagation execution summary.

    iterNum

    The total number of iterations, after which the EP algorithm has reached the convergence criteria

    msgNum

    The total number of messages sent, after which the EP algorithm has reached the convergence criteria

  2. case class ForwardBackwardEPCalibrate(factorGraph: FactorGraph, threshold: Double = 0.00001) extends Product with Serializable

    Calibrates factor graph with forward-backward passes over all factor and variable nodes.

    Calibrates factor graph with forward-backward passes over all factor and variable nodes.

    factorGraph

    Factor graph to be calibrated

    threshold

    Calibration criteria: the maximum absolute difference between old and new corresponding messages on a factor graph

Ungrouped