Class InnerProductOpBase

Provides outgoing messages for InnerProduct(Vector, Vector), given random arguments to the function.

Inheritance

Object

InnerProductOpBase

InnerProductOp

InnerProductOp_PointB

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

public class InnerProductOpBase

Methods

AAverageConditional(Gaussian, Vector, VectorGaussian)

EP message to a.

Declaration

public static VectorGaussian AAverageConditional(Gaussian innerProduct, Vector B, VectorGaussian result)

Parameters

Type	Name	Description
Gaussian	innerProduct	Incoming message from innerProduct. Must be a proper distribution. If uniform, the result will be uniform.
Vector	B	Constant value for b.
VectorGaussian	result	Modified to contain the outgoing message.

Returns

Type	Description
VectorGaussian	result

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_(innerProduct) p(innerProduct) factor(innerProduct,a,b)]/p(a).

Exceptions

Type	Condition
ImproperMessageException	innerProduct is not a proper distribution.

AAverageConditional(Gaussian, Vector, DenseVector, PositiveDefiniteMatrix, VectorGaussian)

EP message to a.

Declaration

public static VectorGaussian AAverageConditional(Gaussian innerProduct, Vector A, DenseVector BMean, PositiveDefiniteMatrix BVariance, VectorGaussian result)

Parameters

Type	Name	Description
Gaussian	innerProduct	Incoming message from innerProduct. Must be a proper distribution. If uniform, the result will be uniform.
Vector	A	Constant value for a.
DenseVector	BMean	Buffer BMean.
PositiveDefiniteMatrix	BVariance	Buffer BVariance.
VectorGaussian	result	Modified to contain the outgoing message.

Returns

Type	Description
VectorGaussian	result

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_(innerProduct) p(innerProduct) factor(innerProduct,a,b)]/p(a).

Exceptions

Type	Condition
ImproperMessageException	innerProduct is not a proper distribution.

AAverageConditional(Double, Vector, VectorGaussian)

EP message to a.

Declaration

[NotSupported("A InnerProduct factor with fixed output is not yet implemented.")]
public static VectorGaussian AAverageConditional(double innerProduct, Vector B, VectorGaussian result)

Parameters

Type	Name	Description
Double	innerProduct	Constant value for innerProduct.
Vector	B	Constant value for b.
VectorGaussian	result	Modified to contain the outgoing message.

Returns

Type	Description
VectorGaussian	result

Remarks

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

AAverageLogarithm(Gaussian, VectorGaussian, DenseVector, PositiveDefiniteMatrix, VectorGaussian)

VMP message to a.

Declaration

public static VectorGaussian AAverageLogarithm(Gaussian innerProduct, VectorGaussian B, DenseVector BMean, PositiveDefiniteMatrix BVariance, VectorGaussian result)

Parameters

Type	Name	Description
Gaussian	innerProduct	Incoming message from innerProduct. Must be a proper distribution. If uniform, the result will be uniform.
VectorGaussian	B	Incoming message from b. Must be a proper distribution. If any element is uniform, the result will be uniform.
DenseVector	BMean	Buffer BMean.
PositiveDefiniteMatrix	BVariance	Buffer BVariance.
VectorGaussian	result	Modified to contain the outgoing message.

Returns

Type	Description
VectorGaussian	result

Remarks

The outgoing message is the exponential of the average log-factor value, where the average is over all arguments except a. Because the factor is deterministic, innerProduct is integrated out before taking the logarithm. The formula is exp(sum_(b) p(b) log(sum_innerProduct p(innerProduct) factor(innerProduct,a,b))).

Exceptions

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

AAverageLogarithm(Gaussian, Vector, VectorGaussian)

VMP message to a.

Declaration

public static VectorGaussian AAverageLogarithm(Gaussian innerProduct, Vector B, VectorGaussian result)

Parameters

Type	Name	Description
Gaussian	innerProduct	Incoming message from innerProduct. Must be a proper distribution. If uniform, the result will be uniform.
Vector	B	Constant value for b.
VectorGaussian	result	Modified to contain the outgoing message.

Returns

Type	Description
VectorGaussian	result

Remarks

The outgoing message is the factor viewed as a function of a with innerProduct integrated out. The formula is sum_innerProduct p(innerProduct) factor(innerProduct,a,b).

Exceptions

Type	Condition
ImproperMessageException	innerProduct is not a proper distribution.

AAverageLogarithm(Double, VectorGaussian, VectorGaussian)

VMP message to a.

Declaration

[NotSupported("Variational Message Passing does not support an InnerProduct factor with fixed output.")]
public static VectorGaussian AAverageLogarithm(double innerProduct, VectorGaussian B, VectorGaussian result)

Parameters

Type	Name	Description
Double	innerProduct	Constant value for innerProduct.
VectorGaussian	B	Incoming message from b. Must be a proper distribution. If any element is uniform, the result will be uniform.
VectorGaussian	result	Modified to contain the outgoing message.

Returns

Type	Description
VectorGaussian	result

Remarks

The outgoing message is the exponential of the average log-factor value, where the average is over all arguments except a. The formula is exp(sum_(b) p(b) log(factor(innerProduct,a,b))).

Exceptions

Type	Condition
ImproperMessageException	B is not a proper distribution.

AAverageLogarithm(Double, Vector, VectorGaussian)

VMP message to a.

Declaration

[NotSupported("A InnerProduct factor with fixed output is not yet implemented.")]
public static VectorGaussian AAverageLogarithm(double innerProduct, Vector B, VectorGaussian result)

Parameters

Type	Name	Description
Double	innerProduct	Constant value for innerProduct.
Vector	B	Constant value for b.
VectorGaussian	result	Modified to contain the outgoing message.

Returns

Type	Description
VectorGaussian	result

Remarks

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

AMean(VectorGaussian, PositiveDefiniteMatrix, DenseVector)

Update the buffer AMean.

Declaration

public static DenseVector AMean(VectorGaussian A, PositiveDefiniteMatrix AVariance, DenseVector result)

Parameters

Type	Name	Description
VectorGaussian	A	Incoming message from a. Must be a proper distribution. If any element is uniform, the result will be uniform.
PositiveDefiniteMatrix	AVariance	Buffer AVariance.
DenseVector	result	Modified to contain the outgoing message.

Returns

Type	Description
DenseVector	result

Remarks

Exceptions

Type	Condition
ImproperMessageException	A is not a proper distribution.

AMeanInit(VectorGaussian)

Initialize the buffer AMean.

Declaration

public static DenseVector AMeanInit(VectorGaussian A)

Parameters

Type	Name	Description
VectorGaussian	A	Incoming message from a.

Returns

Type	Description
DenseVector	Initial value of buffer AMean.

Remarks

AVariance(VectorGaussian, PositiveDefiniteMatrix)

Update the buffer AVariance.

Declaration

public static PositiveDefiniteMatrix AVariance(VectorGaussian A, PositiveDefiniteMatrix result)

Parameters

Type	Name	Description
VectorGaussian	A	Incoming message from a. Must be a proper distribution. If any element is uniform, the result will be uniform.
PositiveDefiniteMatrix	result	Modified to contain the outgoing message.

Returns

Type	Description
PositiveDefiniteMatrix	result

Remarks

Exceptions

Type	Condition
ImproperMessageException	A is not a proper distribution.

AVarianceInit(VectorGaussian)

Initialize the buffer AVariance.

Declaration

public static PositiveDefiniteMatrix AVarianceInit(VectorGaussian A)

Parameters

Type	Name	Description
VectorGaussian	A	Incoming message from a.

Returns

Type	Description
PositiveDefiniteMatrix	Initial value of buffer AVariance.

Remarks

AverageLogFactor()

Evidence message for VMP.

Declaration

public static double AverageLogFactor()

Returns

Type	Description
Double	Zero.

Remarks

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

BAverageConditional(Gaussian, Vector, VectorGaussian)

EP message to b.

Declaration

public static VectorGaussian BAverageConditional(Gaussian innerProduct, Vector A, VectorGaussian result)

Parameters

Type	Name	Description
Gaussian	innerProduct	Incoming message from innerProduct. Must be a proper distribution. If uniform, the result will be uniform.
Vector	A	Constant value for a.
VectorGaussian	result	Modified to contain the outgoing message.

Returns

Type	Description
VectorGaussian	result

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_(innerProduct) p(innerProduct) factor(innerProduct,a,b)]/p(b).

Exceptions

Type	Condition
ImproperMessageException	innerProduct is not a proper distribution.

BAverageConditional(Double, Vector, VectorGaussian)

EP message to b.

Declaration

[NotSupported("A InnerProduct factor with fixed output is not yet implemented.")]
public static VectorGaussian BAverageConditional(double innerProduct, Vector A, VectorGaussian result)

Parameters

Type	Name	Description
Double	innerProduct	Constant value for innerProduct.
Vector	A	Constant value for a.
VectorGaussian	result	Modified to contain the outgoing message.

Returns

Type	Description
VectorGaussian	result

Remarks

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

BAverageLogarithm(Gaussian, VectorGaussian, DenseVector, PositiveDefiniteMatrix, VectorGaussian)

VMP message to b.

Declaration

public static VectorGaussian BAverageLogarithm(Gaussian innerProduct, VectorGaussian A, DenseVector AMean, PositiveDefiniteMatrix AVariance, VectorGaussian result)

Parameters

Type	Name	Description
Gaussian	innerProduct	Incoming message from innerProduct. Must be a proper distribution. If uniform, the result will be uniform.
VectorGaussian	A	Incoming message from a. Must be a proper distribution. If any element is uniform, the result will be uniform.
DenseVector	AMean	Buffer AMean.
PositiveDefiniteMatrix	AVariance	Buffer AVariance.
VectorGaussian	result	Modified to contain the outgoing message.

Returns

Type	Description
VectorGaussian	result

Remarks

The outgoing message is the exponential of the average log-factor value, where the average is over all arguments except b. Because the factor is deterministic, innerProduct is integrated out before taking the logarithm. The formula is exp(sum_(a) p(a) log(sum_innerProduct p(innerProduct) factor(innerProduct,a,b))).

Exceptions

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

BAverageLogarithm(Gaussian, Vector, VectorGaussian)

VMP message to b.

Declaration

public static VectorGaussian BAverageLogarithm(Gaussian innerProduct, Vector A, VectorGaussian result)

Parameters

Type	Name	Description
Gaussian	innerProduct	Incoming message from innerProduct. Must be a proper distribution. If uniform, the result will be uniform.
Vector	A	Constant value for a.
VectorGaussian	result	Modified to contain the outgoing message.

Returns

Type	Description
VectorGaussian	result

Remarks

The outgoing message is the factor viewed as a function of b with innerProduct integrated out. The formula is sum_innerProduct p(innerProduct) factor(innerProduct,a,b).

Exceptions

Type	Condition
ImproperMessageException	innerProduct is not a proper distribution.

BAverageLogarithm(Double, VectorGaussian, VectorGaussian)

VMP message to b.

Declaration

[NotSupported("Variational Message Passing does not support an InnerProduct factor with fixed output.")]
public static VectorGaussian BAverageLogarithm(double innerProduct, VectorGaussian A, VectorGaussian result)

Parameters

Type	Name	Description
Double	innerProduct	Constant value for innerProduct.
VectorGaussian	A	Incoming message from a. Must be a proper distribution. If any element is uniform, the result will be uniform.
VectorGaussian	result	Modified to contain the outgoing message.

Returns

Type	Description
VectorGaussian	result

Remarks

The outgoing message is the exponential of the average log-factor value, where the average is over all arguments except b. The formula is exp(sum_(a) p(a) log(factor(innerProduct,a,b))).

Exceptions

Type	Condition
ImproperMessageException	A is not a proper distribution.

BAverageLogarithm(Double, Vector, VectorGaussian)

VMP message to b.

Declaration

[NotSupported("A InnerProduct factor with fixed output is not yet implemented.")]
public static VectorGaussian BAverageLogarithm(double innerProduct, Vector A, VectorGaussian result)

Parameters

Type	Name	Description
Double	innerProduct	Constant value for innerProduct.
Vector	A	Constant value for a.
VectorGaussian	result	Modified to contain the outgoing message.

Returns

Type	Description
VectorGaussian	result

Remarks

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

BMean(VectorGaussian, PositiveDefiniteMatrix, DenseVector)

Update the buffer BMean.

Declaration

public static DenseVector BMean(VectorGaussian B, PositiveDefiniteMatrix BVariance, DenseVector result)

Parameters

Type	Name	Description
VectorGaussian	B	Incoming message from b. Must be a proper distribution. If any element is uniform, the result will be uniform.
PositiveDefiniteMatrix	BVariance	Buffer BVariance.
DenseVector	result	Modified to contain the outgoing message.

Returns

Type	Description
DenseVector	result

Remarks

Exceptions

Type	Condition
ImproperMessageException	B is not a proper distribution.

BMeanInit(VectorGaussian)

Initialize the buffer BMean.

Declaration

public static DenseVector BMeanInit(VectorGaussian B)

Parameters

Type	Name	Description
VectorGaussian	B	Incoming message from b.

Returns

Type	Description
DenseVector	Initial value of buffer BMean.

Remarks

BVariance(VectorGaussian, PositiveDefiniteMatrix)

Update the buffer BVariance.

Declaration

public static PositiveDefiniteMatrix BVariance(VectorGaussian B, PositiveDefiniteMatrix result)

Parameters

Type	Name	Description
VectorGaussian	B	Incoming message from b. Must be a proper distribution. If any element is uniform, the result will be uniform.
PositiveDefiniteMatrix	result	Modified to contain the outgoing message.

Returns

Type	Description
PositiveDefiniteMatrix	result

Remarks

Exceptions

Type	Condition
ImproperMessageException	B is not a proper distribution.

BVarianceInit(VectorGaussian)

Initialize the buffer BVariance.

Declaration

public static PositiveDefiniteMatrix BVarianceInit(VectorGaussian B)

Parameters

Type	Name	Description
VectorGaussian	B	Incoming message from b.

Returns

Type	Description
PositiveDefiniteMatrix	Initial value of buffer BVariance.

Remarks

InnerProductAverageConditional(DenseVector, PositiveDefiniteMatrix, Vector)

EP message to innerProduct.

Declaration

public static Gaussian InnerProductAverageConditional(DenseVector AMean, PositiveDefiniteMatrix AVariance, Vector B)

Parameters

Type	Name	Description
DenseVector	AMean	Buffer AMean.
PositiveDefiniteMatrix	AVariance	Buffer AVariance.
Vector	B	Constant value for b.

Returns

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

Remarks

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

InnerProductAverageConditional(Vector, DenseVector, PositiveDefiniteMatrix)

EP message to innerProduct.

Declaration

public static Gaussian InnerProductAverageConditional(Vector A, DenseVector BMean, PositiveDefiniteMatrix BVariance)

Parameters

Type	Name	Description
Vector	A	Constant value for a.
DenseVector	BMean	Buffer BMean.
PositiveDefiniteMatrix	BVariance	Buffer BVariance.

Returns

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

Remarks

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

InnerProductAverageConditionalInit()

Declaration

public static Gaussian InnerProductAverageConditionalInit()

Returns

Type	Description
Gaussian

Remarks

InnerProductAverageLogarithm(DenseVector, PositiveDefiniteMatrix, DenseVector, PositiveDefiniteMatrix)

VMP message to innerProduct.

Declaration

public static Gaussian InnerProductAverageLogarithm(DenseVector AMean, PositiveDefiniteMatrix AVariance, DenseVector BMean, PositiveDefiniteMatrix BVariance)

Parameters

Type	Name	Description
DenseVector	AMean	Buffer AMean.
PositiveDefiniteMatrix	AVariance	Buffer AVariance.
DenseVector	BMean	Buffer BMean.
PositiveDefiniteMatrix	BVariance	Buffer BVariance.

Returns

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

Remarks

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

InnerProductAverageLogarithm(DenseVector, PositiveDefiniteMatrix, Vector)

VMP message to innerProduct.

Declaration

public static Gaussian InnerProductAverageLogarithm(DenseVector AMean, PositiveDefiniteMatrix AVariance, Vector B)

Parameters

Type	Name	Description
DenseVector	AMean	Buffer AMean.
PositiveDefiniteMatrix	AVariance	Buffer AVariance.
Vector	B	Constant value for b.

Returns

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

Remarks

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

InnerProductAverageLogarithm(Vector, DenseVector, PositiveDefiniteMatrix)

VMP message to innerProduct.

Declaration

public static Gaussian InnerProductAverageLogarithm(Vector A, DenseVector BMean, PositiveDefiniteMatrix BVariance)

Parameters

Type	Name	Description
Vector	A	Constant value for a.
DenseVector	BMean	Buffer BMean.
PositiveDefiniteMatrix	BVariance	Buffer BVariance.

Returns

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

Remarks

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

InnerProductAverageLogarithmInit()

Declaration

public static Gaussian InnerProductAverageLogarithmInit()

Returns

Type	Description
Gaussian

Remarks

LogAverageFactor(Gaussian, Gaussian)

Evidence message for EP.

Declaration

public static double LogAverageFactor(Gaussian innerProduct, Gaussian to_innerProduct)

Parameters

Type	Name	Description
Gaussian	innerProduct	Incoming message from innerProduct.
Gaussian	to_innerProduct	Outgoing message to innerProduct.

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_(innerProduct) p(innerProduct) factor(innerProduct,a,b)).

LogAverageFactor(Double, DenseVector, PositiveDefiniteMatrix, Vector)

Evidence message for EP.

Declaration

public static double LogAverageFactor(double innerProduct, DenseVector AMean, PositiveDefiniteMatrix AVariance, Vector B)

Parameters

Type	Name	Description
Double	innerProduct	Constant value for innerProduct.
DenseVector	AMean	Buffer AMean.
PositiveDefiniteMatrix	AVariance	Buffer AVariance.
Vector	B	Constant value for b.

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(innerProduct,a,b)).

LogAverageFactor(Double, Vector, DenseVector, PositiveDefiniteMatrix)

Evidence message for EP.

Declaration

public static double LogAverageFactor(double innerProduct, Vector A, DenseVector BMean, PositiveDefiniteMatrix BVariance)

Parameters

Type	Name	Description
Double	innerProduct	Constant value for innerProduct.
Vector	A	Constant value for a.
DenseVector	BMean	Buffer BMean.
PositiveDefiniteMatrix	BVariance	Buffer BVariance.

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(innerProduct,a,b)).

LogEvidenceRatio(Gaussian, VectorGaussian, Vector)

Evidence message for EP.

Declaration

public static double LogEvidenceRatio(Gaussian innerProduct, VectorGaussian a, Vector B)

Parameters

Type	Name	Description
Gaussian	innerProduct	Incoming message from innerProduct.
VectorGaussian	a	Incoming message from a.
Vector	B	Constant value for b.

Returns

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

Remarks

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

LogEvidenceRatio(Gaussian, Vector, VectorGaussian)

Evidence message for EP.

Declaration

public static double LogEvidenceRatio(Gaussian innerProduct, Vector A, VectorGaussian b)

Parameters

Type	Name	Description
Gaussian	innerProduct	Incoming message from innerProduct.
Vector	A	Constant value for a.
VectorGaussian	b	Incoming message from b.

Returns

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

Remarks

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

LogEvidenceRatio(Double, DenseVector, PositiveDefiniteMatrix, Vector)

Evidence message for EP.

Declaration

public static double LogEvidenceRatio(double innerProduct, DenseVector AMean, PositiveDefiniteMatrix AVariance, Vector B)

Parameters

Type	Name	Description
Double	innerProduct	Constant value for innerProduct.
DenseVector	AMean	Buffer AMean.
PositiveDefiniteMatrix	AVariance	Buffer AVariance.
Vector	B	Constant value for b.

Returns

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

Remarks

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

LogEvidenceRatio(Double, Vector, DenseVector, PositiveDefiniteMatrix)

Evidence message for EP.

Declaration

public static double LogEvidenceRatio(double innerProduct, Vector A, DenseVector BMean, PositiveDefiniteMatrix BVariance)

Parameters

Type	Name	Description
Double	innerProduct	Constant value for innerProduct.
Vector	A	Constant value for a.
DenseVector	BMean	Buffer BMean.
PositiveDefiniteMatrix	BVariance	Buffer BVariance.

Returns

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

Remarks

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