mxPenaltyElasticNet {OpenMx}

mxPenaltyElasticNet

Description

Elastic net regularization

Usage

mxPenaltyElasticNet(
  what,
  name,
  alpha = 0,
  alpha.step = 0.1,
  alpha.max = 1,
  lambda = 0,
  lambda.step = 0.1,
  lambda.max = 0.4,
  alpha.min = NA,
  lambda.min = NA,
  epsilon = 1e-05,
  scale = 1,
  ...,
  hyperparams = c("alpha", "lambda")
)

Arguments

what

A character vector of parameters to regularize

name

Name of the regularizer object

alpha

strength of the mixing parameter to be applied at start (default 0.5). Note that 0 indicates a ridge regression with penalty

\frac{lambda}{2}

, and 1 indicates a LASSO regression with penalty lambda.

alpha.step

alpha step during penalty search (default 0.1)

alpha.max

when to end the alpha search (default 1)

lambda

strength of the penalty to be applied at starting values (default 0)

lambda.step

step function for lambda step (default .01)

lambda.max

end of lambda range (default .4)

alpha.min

beginning of the alpha range (default 0)

lambda.min

beginning of the lambda range (default lambda)

epsilon

how close to zero is zero?

scale

a given parameter is divided by scale before comparison with epsilon

...

Not used. Forces remaining arguments to be specified by name

hyperparams

a character vector of hyperparameter names

Details

Applies elastic net regularization. Elastic net is a weighted combination of ridge and LASSO penalties.