Class DerivedVariableOp

Provides outgoing messages for the following factors:

• DerivedVariable<T>(T, out T)
• DerivedVariableInit<T>(T, T, out T)

, given random arguments to the function.

Inheritance

Object

DerivedVariableOp

Inherited Members

Object.Equals(Object)

Object.Equals(Object, Object)

Object.GetHashCode()

Object.GetType()

Object.MemberwiseClone()

Object.ReferenceEquals(Object, Object)

Object.ToString()

Namespace: Microsoft.ML.Probabilistic.Factors

Assembly: Microsoft.ML.Probabilistic.dll

Syntax

[FactorMethod(typeof(Clone), "DerivedVariable<>", new Type[]{}, Default = true)]
[FactorMethod(typeof(Clone), "DerivedVariableInit<>", new Type[]{}, Default = true)]
[Quality(QualityBand.Preview)]
public static class DerivedVariableOp

Methods

DefAverageConditional<T>(T)

EP message to def.

Declaration

public static T DefAverageConditional<T>(T Use)

Parameters

Type	Name	Description
T	Use	Incoming message from use.

Returns

Type	Description
T	The outgoing EP message to the def argument.

Type Parameters

Name	Description
T	The type of the marginal of the variable.

Remarks

The outgoing message is a distribution matching the moments of def as the random arguments are varied. The formula is proj[p(def) sum_(use) p(use) factor(use,def,marginal)]/p(def).

LogAverageFactor()

Evidence message for EP.

Declaration

public static double LogAverageFactor()

Returns

Type	Description
Double	Logarithm of the factor's average value across the given argument distributions.

Remarks

The formula for the result is log(factor(use,def,marginal)).

LogEvidenceRatio<T>(T)

Evidence message for EP.

Declaration

public static double LogEvidenceRatio<T>(T use)

Parameters

Type	Name	Description
T	use	Incoming message from use.

Returns

Type	Description
Double	Logarithm of the factor's contribution the EP model evidence.

Type Parameters

Name	Description
T

Remarks

The formula for the result is log(sum_(use) p(use) factor(use,def,marginal) / sum_use p(use) messageTo(use)). Adding up these values across all factors and variables gives the log-evidence estimate for EP.

MarginalAverageConditional<T>(T, T, T)

EP message to marginal.

Declaration

[SkipIfAllUniform]
public static T MarginalAverageConditional<T>(T Use, T Def, T result)
    where T : SettableToProduct<T>

Parameters

Type	Name	Description
T	Use	Incoming message from use.
T	Def	Incoming message from def.
T	result	Modified to contain the outgoing message.

Returns

Type	Description
T	result

Type Parameters

Name	Description
T	The type of the marginal of the variable.

Remarks

The outgoing message is a distribution matching the moments of marginal as the random arguments are varied. The formula is proj[p(marginal) sum_(use,def) p(use,def) factor(use,def,marginal)]/p(marginal).

MarginalAverageConditional<TDist, T>(T, TDist)

Declaration

public static TDist MarginalAverageConditional<TDist, T>(T Def, TDist result)
    where TDist : HasPoint<T>

Parameters

Type	Name	Description
T	Def
TDist	result

Returns

Type	Description
TDist

Type Parameters

Name	Description
TDist	The type of the marginal of the variable.
T	The type of the incoming message from Def.

MarginalAverageConditionalInit<T>(T)

Declaration

public static T MarginalAverageConditionalInit<T>(T def)
    where T : ICloneable

Parameters

Type	Name	Description
T	def	Incoming message from def.

Returns

Type	Description
T

Type Parameters

Name	Description
T	The type of the marginal of the variable.

Remarks

UseAverageConditional<T>(T)

EP message to use.

Declaration

public static T UseAverageConditional<T>(T Def)

Parameters

Type	Name	Description
T	Def	Incoming message from def.

Returns

Type	Description
T	The outgoing EP message to the use argument.

Type Parameters

Name	Description
T	The type of the marginal of the variable.

Remarks

The outgoing message is a distribution matching the moments of use as the random arguments are varied. The formula is proj[p(use) sum_(def) p(def) factor(use,def,marginal)]/p(use).

UseAverageConditional<TDist, T>(T, TDist)

Declaration

public static TDist UseAverageConditional<TDist, T>(T Def, TDist result)
    where TDist : HasPoint<T>

Parameters

Type	Name	Description
T	Def
TDist	result

Returns

Type	Description
TDist

Type Parameters

Name	Description
TDist	The type of the marginal of the variable.
T	The type of the incoming message from Def.