bayesnet.jayes.transformation

Class SmoothedFactorDecomposition

java.lang.Object

bayesnet.jayes.transformation.AbstractDecomposition

bayesnet.jayes.transformation.SmoothedFactorDecomposition

All Implemented Interfaces:

IDecompositionStrategy

public class SmoothedFactorDecomposition
extends AbstractDecomposition

Applicable if the conditional probability distribution is, in essence, sparse, but was smoothed (which makes it non-sparse). The decomposition makes this explicit in the model.

 Example:
  {
    { 0.99 0.01 }
    { 0.01 0.99 }
    { 0.01 0.99 }
  }
 
  would decompose to the new basis
  
  {
    { 1 1 }
    { 0 1 }
    { 1 0 }
  }
  
  and 
  {
    { 0.01 0 1 }
    { 0.01 1 0 }
    { 0.01 1 0 }
  }
 

This works better than LatentDeterministicDecomposition in cases where the distribution is less close to being deterministic. A good indicator for when to use it is the size of the dimensions. With large dimensions, it becomes unlikely that the distributions can be represented by just one basis vector. In that case, SmoothedFactorDecomposition works better.

Constructor Summary

Constructors

Constructor and Description
SmoothedFactorDecomposition()

Method Summary

Methods inherited from class bayesnet.jayes.transformation.AbstractDecomposition

decompose

Methods inherited from class java.lang.Object

equals, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Detail

SmoothedFactorDecomposition

public SmoothedFactorDecomposition()