Class GateEnterPartialTwoOp

Provides outgoing messages for EnterPartialTwo<T>(Boolean, Boolean, T, Int32[]), given random arguments to the function.

Inheritance

Object

GateEnterPartialTwoOp

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(Gate), "EnterPartialTwo<>", new Type[]{})]
[Quality(QualityBand.Mature)]
public static class GateEnterPartialTwoOp

Methods

AverageLogFactor()

Evidence message for VMP.

Declaration

public static double AverageLogFactor()

Returns

Type	Description
Double	Zero.

Remarks

The formula for the result is log(factor(enterPartialTwo,case0,case1,value,indices)). Adding up these values across all factors and variables gives the log-evidence estimate for VMP.

Case0AverageConditional(Bernoulli)

EP message to case0.

Declaration

public static Bernoulli Case0AverageConditional(Bernoulli result)

Parameters

Type	Name	Description
Bernoulli	result	Modified to contain the outgoing message.

Returns

Type	Description
Bernoulli	result

Remarks

The outgoing message is the factor viewed as a function of case0 conditioned on the given values.

Case0AverageLogarithm<TValue>(IList<TValue>, TValue, Int32[], Bernoulli)

VMP message to case0.

Declaration

public static Bernoulli Case0AverageLogarithm<TValue>(IList<TValue> enterPartialTwo, TValue value, int[] indices, Bernoulli result)

Parameters

Type	Name	Description
IList<TValue>	enterPartialTwo	Incoming message from enterPartialTwo.
TValue	value	Incoming message from value.
Int32[]	indices	Constant value for indices.
Bernoulli	result	Modified to contain the outgoing message.

Returns

Type	Description
Bernoulli	result

Type Parameters

Name	Description
TValue	The type of the message from value.

Remarks

The outgoing message is the exponential of the average log-factor value, where the average is over all arguments except case0. Because the factor is deterministic, enterPartialTwo is integrated out before taking the logarithm. The formula is exp(sum_(value) p(value) log(sum_enterPartialTwo p(enterPartialTwo) factor(enterPartialTwo,case0,case1,value,indices))).

Case1AverageConditional(Bernoulli)

EP message to case1.

Declaration

public static Bernoulli Case1AverageConditional(Bernoulli result)

Parameters

Type	Name	Description
Bernoulli	result	Modified to contain the outgoing message.

Returns

Type	Description
Bernoulli	result

Remarks

The outgoing message is the factor viewed as a function of case1 conditioned on the given values.

Case1AverageLogarithm<TValue>(IList<TValue>, TValue, Int32[], Bernoulli)

VMP message to case1.

Declaration

public static Bernoulli Case1AverageLogarithm<TValue>(IList<TValue> enterPartialTwo, TValue value, int[] indices, Bernoulli result)

Parameters

Type	Name	Description
IList<TValue>	enterPartialTwo	Incoming message from enterPartialTwo.
TValue	value	Incoming message from value.
Int32[]	indices	Constant value for indices.
Bernoulli	result	Modified to contain the outgoing message.

Returns

Type	Description
Bernoulli	result

Type Parameters

Name	Description
TValue	The type of the message from value.

Remarks

The outgoing message is the exponential of the average log-factor value, where the average is over all arguments except case1. Because the factor is deterministic, enterPartialTwo is integrated out before taking the logarithm. The formula is exp(sum_(value) p(value) log(sum_enterPartialTwo p(enterPartialTwo) factor(enterPartialTwo,case0,case1,value,indices))).

EnterPartialTwoAverageConditional<TValue, TResultList>(TValue, TResultList)

EP message to enterPartialTwo.

Declaration

public static TResultList EnterPartialTwoAverageConditional<TValue, TResultList>(TValue value, TResultList result)
    where TResultList : CanSetAllElementsTo<TValue>

Parameters

Type	Name	Description
TValue	value	Incoming message from value. Must be a proper distribution. If uniform, the result will be uniform.
TResultList	result	Modified to contain the outgoing message.

Returns

Type	Description
TResultList	result

Type Parameters

Name	Description
TValue	The type of the message from value.
TResultList	The type of the outgoing message.

Remarks

The outgoing message is a distribution matching the moments of enterPartialTwo as the random arguments are varied. The formula is proj[p(enterPartialTwo) sum_(value) p(value) factor(enterPartialTwo,case0,case1,value,indices)]/p(enterPartialTwo).

Exceptions

Type	Condition
ImproperMessageException	value is not a proper distribution.

EnterPartialTwoAverageLogarithm<TValue, TResultList>(TValue, TResultList)

VMP message to enterPartialTwo.

Declaration

public static TResultList EnterPartialTwoAverageLogarithm<TValue, TResultList>(TValue value, TResultList result)
    where TResultList : CanSetAllElementsTo<TValue>

Parameters

Type	Name	Description
TValue	value	Incoming message from value. Must be a proper distribution. If uniform, the result will be uniform.
TResultList	result	Modified to contain the outgoing message.

Returns

Type	Description
TResultList	result

Type Parameters

Name	Description
TValue	The type of the message from value.
TResultList	The type of the outgoing message.

Remarks

The outgoing message is a distribution matching the moments of enterPartialTwo as the random arguments are varied. The formula is proj[sum_(value) p(value) factor(enterPartialTwo,case0,case1,value,indices)].

Exceptions

Type	Condition
ImproperMessageException	value is not a proper distribution.

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(enterPartialTwo,case0,case1,value,indices)).

LogEvidenceRatio()

Evidence message for EP.

Declaration

public static double LogEvidenceRatio()

Returns

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

Remarks

The formula for the result is log(factor(enterPartialTwo,case0,case1,value,indices)). Adding up these values across all factors and variables gives the log-evidence estimate for EP.

ValueAverageConditional<TValue>(IList<TValue>, Bernoulli, Bernoulli, TValue, Int32[], TValue)

EP message to value.

Declaration

public static TValue ValueAverageConditional<TValue>([SkipIfAllUniform] IList<TValue> enterPartialTwo, Bernoulli case0, Bernoulli case1, TValue value, int[] indices, TValue result)
    where TValue : ICloneable, SettableToUniform, SettableToProduct<TValue>, SettableToRatio<TValue>, SettableToWeightedSum<TValue>, CanGetLogAverageOf<TValue>

Parameters

Type	Name	Description
IList<TValue>	enterPartialTwo	Incoming message from enterPartialTwo. Must be a proper distribution. If all elements are uniform, the result will be uniform.
Bernoulli	case0	Incoming message from case0.
Bernoulli	case1	Incoming message from case1.
TValue	value	Incoming message from value.
Int32[]	indices	Constant value for indices.
TValue	result	Modified to contain the outgoing message.

Returns

Type	Description
TValue	result

Type Parameters

Name	Description
TValue	The type of the message from value.

Remarks

The outgoing message is a distribution matching the moments of value as the random arguments are varied. The formula is proj[p(value) sum_(enterPartialTwo,case0,case1) p(enterPartialTwo,case0,case1) factor(enterPartialTwo,case0,case1,value,indices)]/p(value).

Exceptions

Type	Condition
ImproperMessageException	enterPartialTwo is not a proper distribution.

ValueAverageConditional<TValue, TDomain>(IList<TValue>, Boolean, Boolean, Int32[], TValue)

EP message to value.

Declaration

public static TValue ValueAverageConditional<TValue, TDomain>([SkipIfAllUniform] IList<TValue> enterPartialTwo, bool case1, bool case2, int[] indices, TValue result)
    where TValue : IDistribution<TDomain>, SettableTo<TValue>

Parameters

Type	Name	Description
IList<TValue>	enterPartialTwo	Incoming message from enterPartialTwo. Must be a proper distribution. If all elements are uniform, the result will be uniform.
Boolean	case1	Constant value for case1.
Boolean	case2
Int32[]	indices	Constant value for indices.
TValue	result	Modified to contain the outgoing message.

Returns

Type	Description
TValue	result

Type Parameters

Name	Description
TValue	The type of the message from value.
TDomain	The type of the variable entering the gate.

Remarks

The outgoing message is a distribution matching the moments of value as the random arguments are varied. The formula is proj[p(value) sum_(enterPartialTwo) p(enterPartialTwo) factor(enterPartialTwo,case0,case1,value,indices)]/p(value).

Exceptions

Type	Condition
ImproperMessageException	enterPartialTwo is not a proper distribution.

ValueAverageLogarithm<TValue>(IList<TValue>, Bernoulli, Bernoulli, Int32[], TValue)

VMP message to value.

Declaration

public static TValue ValueAverageLogarithm<TValue>([SkipIfAllUniform] IList<TValue> enterPartialTwo, Bernoulli case0, Bernoulli case1, int[] indices, TValue result)
    where TValue : ICloneable, SettableToProduct<TValue>, SettableToPower<TValue>

Parameters

Type	Name	Description
IList<TValue>	enterPartialTwo	Incoming message from enterPartialTwo. Must be a proper distribution. If all elements are uniform, the result will be uniform.
Bernoulli	case0	Incoming message from case0.
Bernoulli	case1	Incoming message from case1.
Int32[]	indices	Constant value for indices.
TValue	result	Modified to contain the outgoing message.

Returns

Type	Description
TValue	result

Type Parameters

Name	Description
TValue	The type of the message from value.

Remarks

The outgoing message is the exponential of the average log-factor value, where the average is over all arguments except value. Because the factor is deterministic, enterPartialTwo is integrated out before taking the logarithm. The formula is exp(sum_(case0,case1) p(case0,case1) log(sum_enterPartialTwo p(enterPartialTwo) factor(enterPartialTwo,case0,case1,value,indices))).

Exceptions

Type	Condition
ImproperMessageException	enterPartialTwo is not a proper distribution.