Class GammaRatioOp_Laplace

Provides outgoing messages for Ratio(Double, Double), given random arguments to the function.

Inheritance

Object

GammaRatioOp_Laplace

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(Factor), "Ratio", new Type[]{typeof(double), typeof(double)})]
[Buffers(new string[]{"Q"})]
[Quality(QualityBand.Experimental)]
public static class GammaRatioOp_Laplace

Methods

AAverageConditional(Gamma, Gamma, Gamma, Gamma)

EP message to a.

Declaration

public static Gamma AAverageConditional(Gamma ratio, Gamma A, Gamma B, Gamma q)

Parameters

Type	Name	Description
Gamma	ratio	Incoming message from ratio.
Gamma	A	Incoming message from a.
Gamma	B	Incoming message from b. Must be a proper distribution. If uniform, the result will be uniform.
Gamma	q	Buffer q.

Returns

Type	Description
Gamma	The outgoing EP message to the a argument.

Remarks

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

Exceptions

Type	Condition
ImproperMessageException	B is not a proper distribution.

BAverageConditional(Gamma, Gamma, Gamma, Gamma)

EP message to b.

Declaration

public static Gamma BAverageConditional(Gamma ratio, Gamma A, Gamma B, Gamma q)

Parameters

Type	Name	Description
Gamma	ratio	Incoming message from ratio. Must be a proper distribution. If uniform, the result will be uniform.
Gamma	A	Incoming message from a. Must be a proper distribution. If uniform, the result will be uniform.
Gamma	B	Incoming message from b. Must be a proper distribution. If uniform, the result will be uniform.
Gamma	q	Buffer q.

Returns

Type	Description
Gamma	The outgoing EP message to the b argument.

Remarks

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

Exceptions

Type	Condition
ImproperMessageException	ratio is not a proper distribution.
ImproperMessageException	A is not a proper distribution.
ImproperMessageException	B is not a proper distribution.

LogAverageFactor(Gamma, Gamma, Gamma, Gamma)

Evidence message for EP.

Declaration

public static double LogAverageFactor(Gamma ratio, Gamma A, Gamma B, Gamma q)

Parameters

Type	Name	Description
Gamma	ratio	Incoming message from ratio.
Gamma	A	Incoming message from a.
Gamma	B	Incoming message from b.
Gamma	q	Buffer q.

Returns

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

Remarks

The formula for the result is log(sum_(ratio,a,b) p(ratio,a,b) factor(ratio,a,b)).

LogEvidenceRatio(Gamma, Gamma, Gamma, Gamma, Gamma)

Evidence message for EP.

Declaration

public static double LogEvidenceRatio(Gamma ratio, Gamma A, Gamma B, Gamma to_ratio, Gamma q)

Parameters

Type	Name	Description
Gamma	ratio	Incoming message from ratio.
Gamma	A	Incoming message from a.
Gamma	B	Incoming message from b.
Gamma	to_ratio	Previous outgoing message to ratio.
Gamma	q	Buffer q.

Returns

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

Remarks

The formula for the result is log(sum_(ratio,a,b) p(ratio,a,b) factor(ratio,a,b) / sum_ratio p(ratio) messageTo(ratio)). Adding up these values across all factors and variables gives the log-evidence estimate for EP.

LogEvidenceRatio(Double, Gamma, Gamma, Gamma)

Evidence message for EP.

Declaration

public static double LogEvidenceRatio(double ratio, Gamma A, Gamma B, Gamma q)

Parameters

Type	Name	Description
Double	ratio	Constant value for ratio.
Gamma	A	Incoming message from a.
Gamma	B	Incoming message from b.
Gamma	q	Buffer q.

Returns

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

Remarks

The formula for the result is log(sum_(a,b) p(a,b) factor(ratio,a,b)). Adding up these values across all factors and variables gives the log-evidence estimate for EP.

Q(Gamma, Gamma, Gamma)

Update the buffer Q.

Declaration

public static Gamma Q(Gamma ratio, Gamma A, Gamma B)

Parameters

Type	Name	Description
Gamma	ratio	Incoming message from ratio.
Gamma	A	Incoming message from a. Must be a proper distribution. If uniform, the result will be uniform.
Gamma	B	Incoming message from b. Must be a proper distribution. If uniform, the result will be uniform.

Returns

Type	Description
Gamma	New value of buffer Q.

Remarks

Exceptions

Type	Condition
ImproperMessageException	A is not a proper distribution.
ImproperMessageException	B is not a proper distribution.

QInit()

Initialize the buffer Q.

Declaration

public static Gamma QInit()

Returns

Type	Description
Gamma	Initial value of buffer Q.

Remarks

RatioAverageConditional(Gamma, Gamma, Gamma)

EP message to ratio.

Declaration

public static Gamma RatioAverageConditional(Gamma ratio, Gamma A, Gamma B)

Parameters

Type	Name	Description
Gamma	ratio	Incoming message from ratio.
Gamma	A	Incoming message from a. Must be a proper distribution. If uniform, the result will be uniform.
Gamma	B	Incoming message from b. Must be a proper distribution. If uniform, the result will be uniform.

Returns

Type	Description
Gamma	The outgoing EP message to the ratio argument.

Remarks

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

Exceptions

Type	Condition
ImproperMessageException	A is not a proper distribution.
ImproperMessageException	B is not a proper distribution.