sparseCV_MT: sparseCV_MT: internal cross-validation functions. For...

In sMTL: Sparse Multi-Task Learning

sparseCV_MT: internal cross-validation functions. For internal package use only.

Description

sparseCV_MT: internal cross-validation functions. For internal package use only.

Usage

sparseCV_MT(
  data,
  tune.grid,
  hoso = "hoso",
  method = "L0",
  nfolds = "K",
  juliaFnPath = NA,
  messageInd = FALSE,
  LSitr = 50,
  LSspc = 1,
  maxIter = 2500
)

Arguments

data	Matrix with outcome and design matrix
tune.grid	A data.frame of tuning values
hoso	String specifying tuning type
method	Sting specifying regression method
nfolds	String or integer specifying number of folds
juliaFnPath	String specifying path to Julia binary
messageInd	Boolean for message printing
LSitr	Integer specifying do <LSitr> local search iterations on parameter values where we do actually do LS; NA does no local search
LSspc	Integer specifying number of hyperparameters to conduct local search: conduct local search every <LSspc>^th iteration. NA does no local search
maxIter	Integer specifying max iterations of coordinate descent

Value

A list (S3 class) with elements used for cross validation.

best	A dataframe with the hyperparameters associated with the best prediction performance and summary statistics of performance.
best.1se	A dataframe including optimal hyperparameters according to 1-standard deviation rule.
rmse	A dataframe with prediction performance for hyperparamters in tuning grid for all folds.
avg	A dataframe with average performance at each of the hyperparameters in tuning grid (averaged across tasks).

sMTL documentation built on Feb. 16, 2023, 7:29 p.m.