SgdBinaryTrainerBase<TModel>.OptionsBase 类

定义

public class SgdBinaryTrainerBase<TModel>.OptionsBase : Microsoft.ML.Trainers.TrainerInputBaseWithWeight where TModel : class
type SgdBinaryTrainerBase<'Model (requires 'Model : null)>.OptionsBase = class
    inherit TrainerInputBaseWithWeight
Public Class SgdBinaryTrainerBase(Of TModel).OptionsBase
Inherits TrainerInputBaseWithWeight

类型参数

TModel
继承
派生

构造函数

SgdBinaryTrainerBase<TModel>.OptionsBase()

字段

CheckFrequency

确定在迭代次数方面检查收敛的频率。

ConvergenceTolerance

收敛容差。 如果损失减少的指数移动平均值低于此容差,则算法被视为已聚合并停止。

ExampleWeightColumnName

要用于示例权重的列。

(继承自 TrainerInputBaseWithWeight)
FeatureColumnName

要用于功能的列。

(继承自 TrainerInputBase)
L2Regularization

正则化的 L2 权重。

LabelColumnName

要用于标签的列。

(继承自 TrainerInputBaseWithLabel)
LearningRate

新发币使用的初始学习速率。

NumberOfIterations

通过训练数据集的最大传递数。

NumberOfThreads

一度无锁并行度,由SGG使用。

PositiveInstanceWeight

要应用于正类的权重。 这对于使用不平衡的数据进行训练非常有用。

Shuffle

确定是否对每次训练迭代对数据进行混排。

适用于