optLambdas_mgcvWrap: Find optimal ridge penalties with sequential optimization.

In multiridge: Fast Cross-Validation for Multi-Penalty Ridge Regression

Find optimal ridge penalties with sequential optimization.

Description

Sequentially optimizes a marginal likelihood score w.r.t. ridge penalties for multiple data blocks.

Usage

optLambdas_mgcvWrap(penaltiesinit=NULL, XXblocks,Y, pairing=NULL, model=NULL, reltol=1e-4,
  optmethod1= "SANN", optmethod2 =ifelse(length(penaltiesinit)==1,"Brent", "Nelder-Mead"),
  maxItropt1=10,maxItropt2=25,tracescore=TRUE,fixedseed =TRUE, pref=NULL, fixedpen=NULL,
  sigmasq = 1, opt.sigma=ifelse(model=="linear",TRUE, FALSE))

Arguments

penaltiesinit	Numeric vector. Initial values for penaltyparameters. May be obtained from fastCV2.
XXblocks	List of nxn matrices. Usually output of createXXblocks.
Y	Response vector: numeric, binary, factor or survival.
pairing	Numerical vector of length 3 or NULL when pairs are absent. Represents the indices (in XXblocks) of the two data blocks involved in pairing, plus the index of the paired block.
model	Character. Any of c("linear", "logistic", "cox"). Is inferred from Y when NULL.
reltol	Scalar. Relative tolerance for optimization methods.
optmethod1	Character. First, global search method. Any of the methods c("Brent", "Nelder-Mead", "Sann") may be used, but simulated annealing by "Sann" is recommended to search a wide landscape. Other unconstrained methods offered by optim may also be used, but have not been tested.
optmethod2	Character. Second, local search method. Any of the methods c("Brent", "Nelder-Mead", "Sann") may be used, but "Nelder-Mead" is generally recommended. Other unconstrained methods offered by optim may also be used, but have not been tested.
maxItropt1	Integer. Maximum number of iterations for optmethod1.
maxItropt2	Integer. Maximum number of iterations for optmethod2.
tracescore	Boolean. Should the output of the scores be traced?
fixedseed	Boolean. Should the initialization be fixed? For reproducibility.
pref	Integer vector or NULL. Contains indices of data types in XXblocks that are preferential.
fixedpen	Integer vector or NULL. Contains indices of data types of which penalty is fixed to the corresponding value in penaltiesinit.
sigmasq	Default error variance.
opt.sigma	Boolean. Should the error variance be optimized as well? Only relevant for model="linear".

Details

As opposed to optLambdas_mgcv this function first searches globally, then locally. Hence, more time-consuming, but better guarded against multiple local optima. See gam for details on how the marginal likelihood is computed.

Value

List, with components:

res	Outputs of all optimizers used
lambdas	List of penalties found by the optimizers
optpen	Numerical vector with final, optimal penalties

References

Wood, S. N. (2011), Fast stable restricted maximum likelihood and marginal likelihood estimation of semiparametric generalized linear models, J. Roy. Statist. Soc., B 73(1), 3-36.

See Also

optLambdas_mgcv for one-pass optimization. A full demo and data are available from:
https://drive.google.com/open?id=1NUfeOtN8-KZ8A2HZzveG506nBwgW64e4

multiridge documentation built on June 13, 2022, 5:07 p.m.