methodSR: Regularization criterion types

In julien-gibaud/FactorsSCGLR: Supervised Component-based Generalized Linear Regression with Factors

Regularization criterion types

Description

Regularization criterion types

Usage

methodSR(
  phi = "vpi",
  l = 1,
  s = 1/2,
  maxiter = 1000,
  epsilon = 1e-06,
  bailout = 10
)

Arguments

phi	character string describing structural relevance used in the regularization process. Allowed values are "vpi" for Variable Powered Inertia and "cv" for Component Variance. Default to "vpi".
l	is an integer argument (>=1) tuning the importance of variable bundle locality.
s	is a numeric argument (between 0 and 1) tuning the strength of structural relevance with respect to goodness of fit.
maxiter	integer for maximum number of iterations of SR function
epsilon	positive convergence threshold
bailout	integer argument

julien-gibaud/FactorsSCGLR documentation built on March 27, 2024, 8:17 a.m.