MLX-backed elastic net regression — mlxs

Fit lasso or elastic-net penalised regression paths using MLX tensors for the heavy linear algebra. Currently supports Gaussian and binomial families with an optional intercept and column standardisation.

Usage

mlxs_glmnet(
  x,
  y,
  family = mlxs_gaussian(),
  alpha = 1,
  lambda = NULL,
  nlambda = 100,
  lambda_min_ratio = 1e-04,
  standardize = TRUE,
  intercept = TRUE,
  maxit = 1000,
  tol = 1e-06
)

Arguments

x: Numeric matrix of predictors (observations in rows).
y: Numeric response vector (for binomial, values must be 0/1).
family: MLX-aware family object, e.g. mlxs_gaussian() or mlxs_binomial().
alpha: Elastic-net mixing parameter (1 = lasso, currently alpha must be strictly positive).
lambda: Optional decreasing sequence of penalty values. If NULL, a sequence of length nlambda is generated from lambda_max down to lambda_max * lambda_min_ratio.
nlambda: Length of automatically generated lambda path.
lambda_min_ratio: Smallest lambda as a fraction of lambda_max.
standardize: Should columns of x be centred and scaled before fitting? Coefficients are returned on the original scale regardless.
intercept: Should an intercept be fit?
maxit: Maximum proximal-gradient iterations per lambda value.
tol: Convergence tolerance on the coefficient updates.

Value

An object of class mlxs_glmnet containing the fitted coefficient path, intercepts, lambda sequence, and scaling information.

Note

This function is a proof-of-concept. On large dense problems it is typically several times slower than the highly optimised glmnet::glmnet() implementation.