Interface IGeneratedAlgorithm

Interface for running a generated inference algorithm

Namespace: Microsoft.ML.Probabilistic

Assembly: Microsoft.ML.Probabilistic.dll

Syntax

public interface IGeneratedAlgorithm

Properties

NumberOfIterationsDone

The number of iterations done from the initial state or since the last change of observed values

Declaration

int NumberOfIterationsDone { get; }

Property Value

Type	Description
Int32

Remarks

Changing an observed value will reset this to 0.

Methods

Execute(Int32)

Execute the inference algorithm for the specified number of iterations, starting from the initial state

Declaration

void Execute(int numberOfIterations)

Parameters

Type	Name	Description
Int32	numberOfIterations	The number of iterations to perform from the initial state

Remarks

Sets NumberOfIterationsDone = numberOfIterations. This method is equivalent to calling Reset() followed by Update(numberOfIterations).

GetObservedValue(String)

Gets an observed value

Declaration

object GetObservedValue(string variableName)

Parameters

Type	Name	Description
String	variableName	Name of the variable in the generated code

Returns

Type	Description
Object	The observed value

Marginal(String)

Get the marginal distribution (computed up to this point) of a variable

Declaration

object Marginal(string variableName)

Parameters

Type	Name	Description
String	variableName	Name of the variable in the generated code

Returns

Type	Description
Object	The marginal distribution computed up to this point

Remarks

Execute, Update, or Reset must be called first to set the value of the marginal.

Marginal(String, String)

Get the query-specific marginal distribution of a variable. For example, GibbsSampling answers "Marginal", "Samples", and "Conditionals" queries

Declaration

object Marginal(string variableName, string query)

Parameters

Type	Name	Description
String	variableName	Name of the variable in the generated code
String	query	Query string

Returns

Type	Description
Object	The query-specific marginal distribution computed up to this point

Remarks

Execute, Update, or Reset must be called first to set the value of the marginal.

Marginal<T>(String)

Get the marginal distribution (computed up to this point) of a variable, converted to type T

Declaration

T Marginal<T>(string variableName)

Parameters

Type	Name	Description
String	variableName	Name of the variable in the generated code

Returns

Type	Description
T	The marginal distribution computed up to this point

Type Parameters

Name	Description
T	The distribution type.

Remarks

Execute, Update, or Reset must be called first to set the value of the marginal. The conversion operation may be costly, even if the marginal already has type T. For maximum efficiency, use the non-generic Marginal method when conversion is not needed.

Marginal<T>(String, String)

Get the query-specific marginal distribution of a variable, converted to type T

Declaration

T Marginal<T>(string variableName, string query)

Parameters

Type	Name	Description
String	variableName	Name of the variable in the generated code
String	query	Query

Returns

Type	Description
T	The query-specific marginal distribution computed up to this point

Type Parameters

Name	Description
T	Type to cast to

Remarks

Execute, Update, or Reset must be called first to set the value of the marginal. The conversion operation may be costly, even if the marginal already has type T. For maximum efficiency, use the non-generic Marginal method when conversion is not needed.

Reset()

Reset all messages to their initial values. Sets NumberOfIterationsDone to 0

Declaration

void Reset()

Remarks

This method is equivalent to calling Execute(0)

SetObservedValue(String, Object)

Sets an observed value

Declaration

void SetObservedValue(string variableName, object value)

Parameters

Type	Name	Description
String	variableName	Name of the variable in the generated code
Object	value	The observed value

Update(Int32)

Perform additional iterations of the inference algorithm

Declaration

void Update(int additionalIterations)

Parameters

Type	Name	Description
Int32	additionalIterations	The number of additional iterations to perform

Remarks

If no observed values have changed, this method increments NumberOfIterationsDone by additionalIterations, and is equivalent to calling Execute(NumberOfIterationsDone + additionalIterations). If some observed values have changed, this method sets NumberOfIterationsDone = additionalIterations, and is not equivalent to calling Execute, because it will start from the existing message state rather than the initial state.

Events

ProgressChanged

Fired when the progress of inference changes, typically at the end of one iteration of the inference algorithm

Declaration

event EventHandler<ProgressChangedEventArgs> ProgressChanged

Event Type

Type	Description
EventHandler<ProgressChangedEventArgs>