Class GammaSoftmaxOp

Provides outgoing messages for Softmax(IList<Double>), given random arguments to the function.

Inheritance

Object

GammaSoftmaxOp

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(MMath), "Softmax", new Type[]{typeof(IList<double>)})]
[Quality(QualityBand.Experimental)]
public static class GammaSoftmaxOp

Remarks

Here the marginal prototype for logOdds is Gamma, which allows for heavier tailed distributions. This implementation uses the Taylor series bound from Blei 06, followed by optimization using LBFGS. This approach is linear in the dimension K.

Methods

GetMeanLog(IList<Gamma>)

Declaration

public static Vector GetMeanLog(IList<Gamma> x)

Parameters

Type	Name	Description
IList<Gamma>	x	Incoming message from 'x'.

Returns

Type	Description
Vector

Remarks

SoftmaxAverageLogarithm(IList<Gamma>)

VMP message to softmax.

Declaration

public static Dirichlet SoftmaxAverageLogarithm([SkipIfAllUniform] IList<Gamma> x)

Parameters

Type	Name	Description
IList<Gamma>	x	Incoming message from x. Must be a proper distribution. If all elements are uniform, the result will be uniform.

Returns

Type	Description
Dirichlet	The outgoing VMP message to the softmax argument.

Remarks

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

Exceptions

Type	Condition
ImproperMessageException	x is not a proper distribution.

XAverageLogarithm<GammaList>(Dirichlet, GammaList, GammaList)

VMP message to x.

Declaration

public static GammaList XAverageLogarithm<GammaList>(Dirichlet softmax, [SkipIfAllUniform] GammaList x, GammaList result)
    where GammaList : IList<Gamma>

Parameters

Type	Name	Description
Dirichlet	softmax	Incoming message from softmax. Must be a proper distribution. If any element is uniform, the result will be uniform.
GammaList	x	Incoming message from x. Must be a proper distribution. If all elements are uniform, the result will be uniform.
GammaList	result	Modified to contain the outgoing message.

Returns

Type	Description
GammaList	result

Type Parameters

Name	Description
GammaList	The type of the incoming message from x and the outgoing message.

Remarks

The outgoing message is the factor viewed as a function of x with softmax integrated out. The formula is sum_softmax p(softmax) factor(softmax,x).

Exceptions

Type	Condition
ImproperMessageException	softmax is not a proper distribution.
ImproperMessageException	x is not a proper distribution.