generate.lambdas: generate.lambdas

In itsrainingdata/ccdr: CCDr Algorithm for Learning Sparse Gaussian Bayesian Networks.

Description

Convenience function for creating a grid of lambdas.

Usage

generate.lambdas(lambda.max, lambdas.ratio = 0.001, lambdas.length = 50,
  scale = "linear")

Arguments

lambda.max	Maximum value of lambda; in terms of the algorithm this is the initial value of the regularization parameter in the solution path.
lambdas.ratio	Ratio between the maximum lambda value and the minimum lambda value in the solution path.
lambdas.length	Number of values to include.
scale	Which scale to use: Either "linear" or "log".

Details

See Section 5.3 of Aragam and Zhou (2015), JMLR for a discussion of regularization paths (also, solution paths) as used in the CCDr algorithm.

itsrainingdata/ccdr documentation built on May 18, 2019, 7:12 a.m.