MatrixFactorizationModelParameters Class

Definition

Namespace:

Microsoft.ML.Trainers.Recommender

Assembly:

Microsoft.ML.Recommender.dll

Package:

Microsoft.ML.Recommender v0.21.1

Model parameters for MatrixFactorizationTrainer.

public sealed class MatrixFactorizationModelParameters : Microsoft.ML.ICanSaveModel

type MatrixFactorizationModelParameters = class
    interface ICanSaveModel

Public NotInheritable Class MatrixFactorizationModelParameters
Implements ICanSaveModel

Inheritance

Object

MatrixFactorizationModelParameters

Implements

ICanSaveModel

Remarks

MatrixFactorizationModelParameters stores two factor matrices, P and Q, for approximating the training matrix, R, by P * Q, where * is a matrix multiplication. This model expects two inputs, row index and column index, and produces the (approximated) value at the location specified by the two inputs in R. More specifically, if input row and column indices are u and v, respectively. The output (a scalar) would be the inner product of the u-th row in P and the v-th column in Q.

Fields

ApproximationRank	The rank of the factor matrices.
NumberOfColumns	The number of columns.
NumberOfRows	The number of rows.

Properties

LeftFactorMatrix	Left approximation matrix
RightFactorMatrix	Right approximation matrix

Explicit Interface Implementations

ICanSaveModel.Save(ModelSaveContext)

Save model to the given context