huge_select

Usage

huge_select(
    est,
    criterion=None,
    ebic_gamma=0.5,
    stars_thresh=0.1,
    stars_subsample_ratio=None,
    rep_num=20,
    verbose=True,
    backend="native",
) -> HugeSelectResult

Description

Native model selection for an estimated graph path.

Key arguments

• est: HugeResult
• criterion: "ric", "stars", or "ebic"
• ebic_gamma: EBIC tuning parameter
• stars_thresh: threshold in (0, 1]
• stars_subsample_ratio: optional subsample ratio
• rep_num: repetition count for stochastic criteria
• backend: currently only "native"

Returns

HugeSelectResult with opt_lambda, opt_sparsity, refit, and optional fields (opt_icov, opt_cov, variability, ebic_score).

Notes

• criterion="ebic" requires a glasso fit.
• opt_index is 1-based for compatibility with prior wrapper behavior.