VineForestDensity.__init__

VineForestDensity.__init__(base_params=None, n_vines=100, vines_sampling='uniform', bootstrap=True, val_fraction=0.25, best_only=False, method='da_mcs_marg', alpha=0.05, add_dissmann=True, random_state=None, n_jobs=1, verbose=False)

Ensemble of vine-copula density estimators with random structures.

Builds n_vines VineDensity base learners on randomly sampled vine structures, prunes them via the model confidence set (MCS), and averages the per-sample density values at predict time.

Parameters:

base_paramsdict or None, default=None

Keyword arguments forwarded to each VineDensity __init__. Example: {"batch_size": 200}.

n_vinesint, default=100

Number of random base estimators before MCS pruning.

vines_sampling{“uniform”, “local”}, default=”uniform”

Random-structure generator. "uniform" draws uniformly over R-vines via Joe’s algorithm; "local" draws each tree from the Kendall’s-tau-weighted distribution (Dissmann MST as mode) via Wilson’s loop-erased random walk.

bootstrapbool, default=True

Bootstrap-resample the training set for each base estimator.

val_fractionfloat, default=0.25

Held-out fraction used for MCS survivor selection. 0 disables validation.

best_onlybool, default=False

Keep only the single best survivor rather than the full MCS.

method{“da_mcs_marg”, “da_mcs_unif”} or None, default=”da_mcs_marg”

Survivor-selection method. None keeps anything strictly better than the Dissmann baseline.

alphafloat, default=0.05

Significance level for the MCS selector.

add_dissmannbool, default=True

Include the Dissmann-structure baseline among candidates.

random_stateint, RandomState instance or None, default=None

Seed for reproducibility.

n_jobsint, default=1

Number of joblib workers used during fit and pdf evaluation.

verbosebool, default=False

Warn if no random estimator beats the default.