损失函数:分类和回归损失函数

分类和回归的损失函数。

用法

  expLoss(beta = 1, ...)

  hingeLoss(margin = 1, ...)

  logLoss(...)

  smoothHingeLoss(smoothingConst = 1, ...)

  poissonLoss(...)

  squaredLoss(...)

参数

beta

指定 beta (dilation) 的数值。 默认值为 1。

margin

指定数值边距值。 默认值为 1。

smoothingConst

指定平滑常数的数值。 默认值为 1。

...

隐藏参数。

详细信息

损失函数衡量机器学习算法的预测值与监督的输出之间的差异,并表示出错的代价。

支持的分类损失函数包括:

logLoss

expLoss

hingeLoss

smoothHingeLoss

支持的回归损失函数包括:

poissonLoss

squaredLoss.

定义损失函数的字符串。

作者

Microsoft Corporation Microsoft Technical Support

另请参阅

rxFastLinearrxNeuralNet

示例


 train <- function(lossFunction) {

     result <- rxFastLinear(isCase ~ age + parity + education + spontaneous + induced,
                   transforms = list(isCase = case == 1), lossFunction = lossFunction,
                   data = infert,
                   type = "binary")
     coef(result)[["age"]]
 }

 age <- list()
 age$LogLoss <- train(logLoss())
 age$LogLossHinge <- train(hingeLoss())
 age$LogLossSmoothHinge <- train(smoothHingeLoss())
 age