tune_lasso: tune_lasso

In vviers/DIYLassoElasticNet: Coordinate-Descent for Penalised Regression Problems (ST443 Final Project)

Description

finds optimal lambda, given train and validation sets

Usage

tune_lasso(lambda_max = 10, step_lambda = 0.1, ytrain, Xtrain, yvalid,
  Xvalid)

Arguments

lambda_max	maximum lambda to try
step_lambda	step size between 0 and lambda_max
ytrain	a nx1 label vector used for training
Xtrain	the design matrix used for training
yvalid	a nx1 label vector used for validating
Xvalid	the design matrix used for validating
n_folds	the number of folds to use in cross-validation
one_stderr_rule	whether to use the one-standard-error rule when choosing lambda

Value

a list object containing a data.frame summarizing the error for each pair of lambda/alpha, and the best lambda/alpha

vviers/DIYLassoElasticNet documentation built on May 30, 2019, 12:49 p.m.