Function signature for a function that can compute a performance metric (i.e. a ILossT) from a set of expected (ground-truth) and actual (model prediction) output values. Additional information about the metric (such as its variance) or the learning problem (such as the expected number of classes) can be set in the object passed as the info parameter.

Namespace: Accord.MachineLearning.Performance
Assembly: Accord.MachineLearning (in Accord.MachineLearning.dll) Version: 3.8.0

Syntax

Copy

public delegate double ComputeLoss<TOutput, TInfo>(
	TOutput[] expected,
	TOutput[] actual,
	TInfo info
)

Public Delegate Function ComputeLoss(Of TOutput, TInfo) ( 
	expected As TOutput(),
	actual As TOutput(),
	info As TInfo
) As Double

Parameters

expected: Type: TOutput
The ground-truth data that the model was supposed to predict.
actual: Type: TOutput
The data that the model has actually predicted.
info: Type: TInfo
A info object (e.g. SetResultTModel) that can be used to obtain more information about the data split being evaluated and store additional information about the computed metric.

Type Parameters

TOutput
TInfo

Return Value

Type: Double
A metric that measures how far the model predictions were from the expected ground-truth.

Reference

Accord.MachineLearning.Performance Namespace

ComputeLossAddLanguageSpecificTextSet("LSTACBBA866_0?cs=&lt;|vb=(Of |cpp=&lt;|fs=&lt;'|nu=(");TOutput, TInfoAddLanguageSpecificTextSet("LSTACBBA866_1?cs=&gt;|vb=)|cpp=&gt;|fs=&gt;|nu=)"); Delegate

Parameters

Type Parameters

Return Value

Reference

ComputeLossTOutput, TInfo Delegate