Fit solution paths for linear or logistic regression models penalized by lasso (alpha = 1) or elastic-net (1e-4 < alpha < 1) over a grid of values for the regularization parameter lambda.
COPY_biglasso_main(
X,
y.train,
ind.train,
ind.col,
covar.train,
family = c("gaussian", "binomial"),
alphas = 1,
K = 10,
ind.sets = NULL,
nlambda = 200,
lambda.min.ratio = if (n > p) 1e-04 else 0.001,
nlam.min = 50,
n.abort = 10,
base.train = NULL,
pf.X = NULL,
pf.covar = NULL,
eps = 1e-05,
max.iter = 1000,
dfmax = 50000,
lambda.min = if (n > p) 1e-04 else 0.001,
power_scale = 1,
power_adaptive = 0,
return.all = FALSE,
warn = TRUE,
ncores = 1
)Either "gaussian" (linear) or "binomial" (logistic).
The elastic-net mixing parameter that controls the relative
contribution from the lasso (l1) and the ridge (l2) penalty. The penalty is
defined as $$ \alpha||\beta||_1 + (1-\alpha)/2||\beta||_2^2.$$
alpha = 1 is the lasso penalty and alpha in between 0
(1e-4) and 1 is the elastic-net penalty. Default is 1. You can
pass multiple values, and only one will be used (optimized by grid-search).
Number of sets used in the Cross-Model Selection and Averaging
(CMSA) procedure. Default is 10.
Integer vectors of values between 1 and K specifying
which set each index of the training set is in. Default randomly assigns
these values but it can be useful to set this vector for reproducibility,
or if you want to refine the grid-search over alphas using the same sets.
The number of lambda values. Default is 200.
The smallest value for lambda, as a fraction of
lambda.max. Default is .0001 if the number of observations is larger than
the number of variables and .001 otherwise.
Minimum number of lambda values to investigate. Default is 50.
Number of lambda values for which prediction on the validation
set must decrease before stopping. Default is 10.
Vector of base predictions. Model will be learned starting from these predictions. This can be useful if you want to previously fit a model with large-effect variables that you don't want to penalize.
A multiplicative factor for the penalty applied to each coefficient.
If supplied, pf.X must be a numeric vector of the same length as ind.col.
Default is all 1. The purpose of pf.X is to apply differential
penalization if some coefficients are thought to be more likely than others
to be in the model. Setting SOME to 0 allows to have unpenalized coefficients.
Same as pf.X, but for covar.train.
You might want to set some to 0 as variables with large effects can mask
small effects in penalized regression.
Convergence threshold for inner coordinate descent.
The algorithm iterates until the maximum change in the objective after any
coefficient update is less than eps times the null deviance.
Default value is 1e-5.
Maximum number of iterations. Default is 1000.
Upper bound for the number of nonzero coefficients. Default is
50e3 because, for large data sets, computational burden may be
heavy for models with a large number of nonzero coefficients.
This parameter has been renamed lambda.min.ratio and is
now deprecated.
When using lasso (alpha = 1), penalization to apply that
is equivalent to scaling genotypes dividing by (standard deviation)^power_scale.
Default is 1 and corresponding to standard scaling. Using 0 would correspond
to using unscaled variables and using 0.5 is Pareto scaling. If you e.g. use
power_scale = c(0, 0.5, 1), the best value in CMSA will be used
(just like with alphas).
Multiplicative penalty factor to apply to variables
in the form of 1 / m_j^power_adaptive, where m_j is the marginal statistic
for variable j. Default is 0, which effectively disables this option.
If you e.g. use power_adaptive = c(0, 0.5, 1.5), the best value in CMSA
will be used (just like with alphas).
Deprecated. Now always return all models.
Whether to warn if some models may not have reached a minimum.
Default is TRUE.
The objective function for linear regression (family = "gaussian") is
$$\frac{1}{2n}\textrm{RSS} + \textrm{penalty},$$ for logistic regression
(family = "binomial") it is $$-\frac{1}{n} loglike +
\textrm{penalty}.$$