ComputeLogisticRegressionStandardDeviation.ComputeStandardDeviation(Double[], Int32[], Int32, Int32, IChannel, Single) Method

Definition

Computes the standard deviation matrix of each of the non-zero training weights, needed to calculate further the standard deviation, p-value and z-Score. The calculations are not part of Microsoft.ML package, due to the size of MKL. If you need these calculations, add the Microsoft.ML.Mkl.Components package, and initialize ComputeStandardDeviation to the ComputeLogisticRegressionStandardDeviation implementation in the Microsoft.ML.Mkl.Components package. Due to the existence of regularization, an approximation is used to compute the variances of the trained linear coefficients.

public abstract Microsoft.ML.Data.VBuffer<float> ComputeStandardDeviation (double[] hessian, int[] weightIndices, int parametersCount, int currentWeightsCount, Microsoft.ML.Runtime.IChannel ch, float l2Weight);
abstract member ComputeStandardDeviation : double[] * int[] * int * int * Microsoft.ML.Runtime.IChannel * single -> Microsoft.ML.Data.VBuffer<single>
Public MustOverride Function ComputeStandardDeviation (hessian As Double(), weightIndices As Integer(), parametersCount As Integer, currentWeightsCount As Integer, ch As IChannel, l2Weight As Single) As VBuffer(Of Single)

Parameters

hessian
Double[]
weightIndices
Int32[]
parametersCount
Int32
currentWeightsCount
Int32
l2Weight
Single

Returns

Applies to