Class LogisticOp_SJ99

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

Inheritance

Object

LogisticOp_SJ99

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), "Logistic", new Type[]{typeof(double)}, Default = true)]
[Quality(QualityBand.Preview)]
[Buffers(new string[]{"A"})]
public class LogisticOp_SJ99

Remarks

Uses the Saul and Jordan (1999) bound \langle log(1+exp(x)) \rangle \leq a^2*v/2 + log(1+exp(m+(1-2a)v/2))

Fields

global_step

Declaration

public static double global_step

Field Value

Type	Description
Double

Methods

A(Gaussian, Double)

Update the buffer A.

Declaration

public static double A(Gaussian x, double a)

Parameters

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

Returns

Type	Description
Double	New value of buffer A.

Remarks

Exceptions

Type	Condition
ImproperMessageException	x is not a proper distribution.

AInit()

Initialize the buffer A.

Declaration

public static double AInit()

Returns

Type	Description
Double	Initial value of buffer A.

Remarks

AverageLogFactor(Beta, Gaussian, Beta, Double)

Evidence message for VMP.

Declaration

public static double AverageLogFactor(Beta logistic, Gaussian x, Beta to_logistic, double a)

Parameters

Type	Name	Description
Beta	logistic	Incoming message from logistic.
Gaussian	x	Incoming message from x. Must be a proper distribution. If uniform, the result will be uniform.
Beta	to_logistic	Previous outgoing message to logistic.
Double	a	Buffer a.

Returns

Type	Description
Double	Zero.

Remarks

In Variational Message Passing, the evidence contribution of a deterministic factor is zero. Adding up these values across all factors and variables gives the log-evidence estimate for VMP.

Exceptions

Type	Condition
ImproperMessageException	x is not a proper distribution.

LogisticAverageLogarithm(Gaussian, Double)

VMP message to logistic.

Declaration

public static Beta LogisticAverageLogarithm(Gaussian x, double a)

Parameters

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

Returns

Type	Description
Beta	The outgoing VMP message to the logistic argument.

Remarks

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

Exceptions

Type	Condition
ImproperMessageException	x is not a proper distribution.

LogisticInit()

Initialize the buffer logistic.

Declaration

public static Beta LogisticInit()

Returns

Type	Description
Beta	Initial value of buffer logistic.

Remarks

logSumExpBound(Double, Double, Double)

Declaration

public static double logSumExpBound(double m, double v, double a)

Parameters

Type	Name	Description
Double	m
Double	v
Double	a

Returns

Type	Description
Double

XAverageLogarithm(Beta, Gaussian, Gaussian, Double)

VMP message to x.

Declaration

public static Gaussian XAverageLogarithm(Beta logistic, Gaussian x, Gaussian to_x, double a)

Parameters

Type	Name	Description
Beta	logistic	Incoming message from logistic. Must be a proper distribution. If uniform, the result will be uniform.
Gaussian	x	Incoming message from x.
Gaussian	to_x	Previous outgoing message to x.
Double	a	Buffer a.

Returns

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

Remarks

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

Exceptions

Type	Condition
ImproperMessageException	logistic is not a proper distribution.

XInit()

Initialize the buffer x.

Declaration

public static Gaussian XInit()

Returns

Type	Description
Gaussian	Initial value of buffer x.

Remarks