Class ExpOp

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

Inheritance

Object

ExpOp

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(Math), "Exp", new Type[]{typeof(double)}, Default = true)]
[Quality(QualityBand.Stable)]
public class ExpOp

Fields

ForceProper

Forces proper messages when set to true.

Declaration

public static bool ForceProper

Field Value

Type	Description
Boolean

QuadratureIterations

Number of quadrature iterations

Declaration

public static int QuadratureIterations

Field Value

Type	Description
Int32

QuadratureNodeCount

The number of quadrature nodes used to compute the messages. Reduce this number to save time in exchange for less accuracy.

Declaration

public static int QuadratureNodeCount

Field Value

Type	Description
Int32

QuadratureShift

Quadrature shift

Declaration

public static bool QuadratureShift

Field Value

Type	Description
Boolean

UseRandomDamping

Declaration

public static bool UseRandomDamping

Field Value

Type	Description
Boolean

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(exp,d)). Adding up these values across all factors and variables gives the log-evidence estimate for VMP.

AverageLogFactor(Double, Double)

Evidence message for VMP.

Declaration

public static double AverageLogFactor(double exp, double d)

Parameters

Type	Name	Description
Double	exp	Constant value for exp.
Double	d	Constant value for d.

Returns

Type	Description
Double	Zero.

Remarks

The formula for the result is log(factor(exp,d)). Adding up these values across all factors and variables gives the log-evidence estimate for VMP.

DAverageConditional(Gamma, Gaussian, Gaussian)

EP message to d.

Declaration

public static Gaussian DAverageConditional(Gamma exp, Gaussian d, Gaussian result)

Parameters

Type	Name	Description
Gamma	exp	Incoming message from exp. Must be a proper distribution. If uniform, the result will be uniform.
Gaussian	d	Incoming message from d. Must be a proper distribution. If uniform, the result will be uniform.
Gaussian	result	Modified to contain the outgoing message.

Returns

Type	Description
Gaussian	result

Remarks

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

Exceptions

Type	Condition
ImproperMessageException	exp is not a proper distribution.
ImproperMessageException	d is not a proper distribution.

DAverageConditional(GammaPower, Gaussian, Gaussian)

Declaration

public static Gaussian DAverageConditional(GammaPower exp, Gaussian d, Gaussian to_d)

Parameters

Type	Name	Description
GammaPower	exp
Gaussian	d
Gaussian	to_d

Returns

Type	Description
Gaussian

DAverageConditional(Double)

EP message to d.

Declaration

public static Gaussian DAverageConditional(double exp)

Parameters

Type	Name	Description
Double	exp	Constant value for exp.

Returns

Type	Description
Gaussian	The outgoing EP message to the d argument.

Remarks

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

DAverageLogarithm(Gamma, Gaussian, Gaussian)

VMP message to d.

Declaration

public static Gaussian DAverageLogarithm(Gamma exp, Gaussian d, Gaussian to_d)

Parameters

Type	Name	Description
Gamma	exp	Incoming message from exp. Must be a proper distribution. If uniform, the result will be uniform.
Gaussian	d	Incoming message from d. Must be a proper distribution. If uniform, the result will be uniform.
Gaussian	to_d	Previous outgoing message to d.

Returns

Type	Description
Gaussian	The outgoing VMP message to the d argument.

Remarks

The outgoing message is the factor viewed as a function of d with exp integrated out. The formula is sum_exp p(exp) factor(exp,d).

Exceptions

Type	Condition
ImproperMessageException	exp is not a proper distribution.
ImproperMessageException	d is not a proper distribution.

DAverageLogarithm(Gamma, NonconjugateGaussian, NonconjugateGaussian)

VMP message to d.

Declaration

[Quality(QualityBand.Experimental)]
public static NonconjugateGaussian DAverageLogarithm(Gamma exp, NonconjugateGaussian d, NonconjugateGaussian result)

Parameters

Type	Name	Description
Gamma	exp	Incoming message from exp. Must be a proper distribution. If uniform, the result will be uniform.
NonconjugateGaussian	d	Incoming message from d. Must be a proper distribution. If uniform, the result will be uniform.
NonconjugateGaussian	result	Modified to contain the outgoing message.

Returns

Type	Description
NonconjugateGaussian	result

Remarks

The outgoing message is the factor viewed as a function of d with exp integrated out. The formula is sum_exp p(exp) factor(exp,d).

Exceptions

Type	Condition
ImproperMessageException	exp is not a proper distribution.
ImproperMessageException	d is not a proper distribution.

DAverageLogarithm(Double)

VMP message to d.

Declaration

public static Gaussian DAverageLogarithm(double exp)

Parameters

Type	Name	Description
Double	exp	Constant value for exp.

Returns

Type	Description
Gaussian	The outgoing VMP message to the d argument.

Remarks

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

DNonconjugateAverageLogarithm(Gamma, Gaussian, NonconjugateGaussian)

Nonconjugate VMP message to 'd'.

Declaration

[Quality(QualityBand.Experimental)]
public static NonconjugateGaussian DNonconjugateAverageLogarithm(Gamma exp, Gaussian d, NonconjugateGaussian result)

Parameters

Type	Name	Description
Gamma	exp	Incoming message from 'exp'. Must be a proper distribution. If uniform, the result will be uniform.
Gaussian	d
NonconjugateGaussian	result

Returns

Type	Description
NonconjugateGaussian	The outgoing nonconjugate VMP message to the 'd' argument.

Remarks

The outgoing message is the exponential of the integral of the log-factor times incoming messages, over all arguments except 'd'. The formula is int log(f(d,x)) q(x) dx where x = (exp).

Exceptions

Type	Condition
ImproperMessageException	exp is not a proper distribution

ExpAverageConditional(Gamma, Gaussian, Gaussian)

EP message to exp.

Declaration

public static Gamma ExpAverageConditional(Gamma exp, Gaussian d, Gaussian to_d)

Parameters

Type	Name	Description
Gamma	exp	Incoming message from exp.
Gaussian	d	Incoming message from d. Must be a proper distribution. If uniform, the result will be uniform.
Gaussian	to_d	Previous outgoing message to d.

Returns

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

Remarks

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

Exceptions

Type	Condition
ImproperMessageException	d is not a proper distribution.

ExpAverageConditional(GammaPower, Gaussian, Gaussian)

EP message to exp.

Declaration

public static GammaPower ExpAverageConditional(GammaPower exp, Gaussian d, Gaussian to_d)

Parameters

Type	Name	Description
GammaPower	exp	Incoming message from exp.
Gaussian	d	Incoming message from d. Must be a proper distribution. If uniform, the result will be uniform.
Gaussian	to_d	Previous outgoing message to d.

Returns

Type	Description
GammaPower	The outgoing EP message to the exp argument.

Remarks

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

Exceptions

Type	Condition
ImproperMessageException	d is not a proper distribution.

ExpAverageConditionalInit(Gaussian)

Declaration

public static Gamma ExpAverageConditionalInit(Gaussian d)

Parameters

Type	Name	Description
Gaussian	d	Incoming message from d.

Returns

Type	Description
Gamma

Remarks

ExpAverageLogarithm(Gaussian)

VMP message to exp.

Declaration

public static Gamma ExpAverageLogarithm(Gaussian d)

Parameters

Type	Name	Description
Gaussian	d	Incoming message from d. Must be a proper distribution. If uniform, the result will be uniform.

Returns

Type	Description
Gamma	The outgoing VMP message to the exp argument.

Remarks

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

Exceptions

Type	Condition
ImproperMessageException	d is not a proper distribution.

ExpAverageLogarithm(Gaussian, GammaPower)

VMP message to exp.

Declaration

public static GammaPower ExpAverageLogarithm(Gaussian d, GammaPower result)

Parameters

Type	Name	Description
Gaussian	d	Incoming message from d. Must be a proper distribution. If uniform, the result will be uniform.
GammaPower	result	Modified to contain the outgoing message.

Returns

Type	Description
GammaPower	result

Remarks

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

Exceptions

Type	Condition
ImproperMessageException	d is not a proper distribution.

ExpAverageLogarithm(NonconjugateGaussian)

VMP message to exp.

Declaration

[Quality(QualityBand.Experimental)]
public static Gamma ExpAverageLogarithm(NonconjugateGaussian d)

Parameters

Type	Name	Description
NonconjugateGaussian	d	Incoming message from d. Must be a proper distribution. If uniform, the result will be uniform.

Returns

Type	Description
Gamma	The outgoing VMP message to the exp argument.

Remarks

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

Exceptions

Type	Condition
ImproperMessageException	d is not a proper distribution.

ExpAverageLogarithmInit()

Declaration

public static Gamma ExpAverageLogarithmInit()

Returns

Type	Description
Gamma

Remarks

LogAverageFactor(CanGetLogProb<Double>, Double)

Evidence message for EP.

Declaration

public static double LogAverageFactor(CanGetLogProb<double> exp, double d)

Parameters

Type	Name	Description
CanGetLogProb<Double>	exp	Incoming message from exp.
Double	d	Constant value for d.

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_(exp) p(exp) factor(exp,d)).

LogAverageFactor(Gamma, Gaussian, Gaussian)

Evidence message for EP.

Declaration

public static double LogAverageFactor(Gamma exp, Gaussian d, Gaussian to_d)

Parameters

Type	Name	Description
Gamma	exp	Incoming message from exp.
Gaussian	d	Incoming message from d.
Gaussian	to_d	Previous outgoing message to d.

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_(exp,d) p(exp,d) factor(exp,d)).

LogAverageFactor(GammaPower, Gaussian, Gaussian)

Evidence message for EP.

Declaration

public static double LogAverageFactor(GammaPower exp, Gaussian d, Gaussian to_d)

Parameters

Type	Name	Description
GammaPower	exp	Incoming message from exp.
Gaussian	d	Incoming message from d.
Gaussian	to_d	Previous outgoing message to d.

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_(exp,d) p(exp,d) factor(exp,d)).

LogAverageFactor(Double, Gaussian)

Evidence message for EP.

Declaration

public static double LogAverageFactor(double exp, Gaussian d)

Parameters

Type	Name	Description
Double	exp	Constant value for exp.
Gaussian	d	Incoming message from d.

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_(d) p(d) factor(exp,d)).

LogAverageFactor(Double, Double)

Evidence message for EP.

Declaration

public static double LogAverageFactor(double exp, double d)

Parameters

Type	Name	Description
Double	exp	Constant value for exp.
Double	d	Constant value for d.

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(exp,d)).

LogAverageFactor_slow(Gamma, Gaussian)

Evidence message for EP

Declaration

public static double LogAverageFactor_slow(Gamma exp, Gaussian d)

Parameters

Type	Name	Description
Gamma	exp	Incoming message from 'exp'.
Gaussian	d	Incoming message from 'd'.

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_(exp,d) p(exp,d) factor(exp,d)).

LogEvidenceRatio(Gamma, Gaussian, Gamma, Gaussian)

Evidence message for EP.

Declaration

public static double LogEvidenceRatio(Gamma exp, Gaussian d, Gamma to_exp, Gaussian to_d)

Parameters

Type	Name	Description
Gamma	exp	Incoming message from exp.
Gaussian	d	Incoming message from d.
Gamma	to_exp	Outgoing message to exp.
Gaussian	to_d	Previous outgoing message to d.

Returns

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

Remarks

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

LogEvidenceRatio(GammaPower, Gaussian, GammaPower, Gaussian)

Evidence message for EP.

Declaration

public static double LogEvidenceRatio(GammaPower exp, Gaussian d, GammaPower to_exp, Gaussian to_d)

Parameters

Type	Name	Description
GammaPower	exp	Incoming message from exp.
Gaussian	d	Incoming message from d.
GammaPower	to_exp	Outgoing message to exp.
Gaussian	to_d	Previous outgoing message to d.

Returns

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

Remarks

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

LogEvidenceRatio(Double, Gaussian)

Evidence message for EP.

Declaration

public static double LogEvidenceRatio(double exp, Gaussian d)

Parameters

Type	Name	Description
Double	exp	Constant value for exp.
Gaussian	d	Incoming message from d.

Returns

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

Remarks

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

LogEvidenceRatio(Double, Double)

Evidence message for EP.

Declaration

public static double LogEvidenceRatio(double exp, double d)

Parameters

Type	Name	Description
Double	exp	Constant value for exp.
Double	d	Constant value for d.

Returns

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

Remarks

The formula for the result is log(factor(exp,d)). Adding up these values across all factors and variables gives the log-evidence estimate for EP.