Parameters

Shared

The core parameters are accepted by all SnapBoost estimators. Adaptive controls are exposed by the recommended SnapBoostClassifier and SnapBoostRegressor classes.

Parameter	Type	Default	Description
num_iterations	int	100	Number of boosting rounds
learning_rate	float	0.1	Shrinkage applied to each learner’s contribution
random_state	int or None	None	Seed for learner selection and tree fitting
verbose	bool	False	Show a tqdm progress bar during training
selection_strategy	{"random", "greedy"}	"random"	Sample one learner or select the lowest-loss candidate
line_search	bool	False	Select a contribution weight per boosting round
subsample	float	1.0	Fraction of rows used to fit each learner
early_stopping_rounds	positive int or None	None	Validation patience before restoring the best ensemble
min_delta	float	0.0	Minimum validation-loss improvement
objective	str	"auto"	Loss to optimize. Classifiers accept "auto" and "log_loss"; regressors also accept "squared_error", "pseudo_huber", and "quantile"
objective_parameter	float or None	None	Pseudo-Huber delta (default 1.0) or quantile level (default 0.5); ignored otherwise

Classifiers select logistic loss for binary targets and softmax for multiclass targets under both objective="auto" and objective="log_loss".

Warning

The Newton working response for pseudo-Huber grows like residual³ / delta², so the default delta=1.0 diverges on targets that are not roughly unit-scale. Standardize y, or set objective_parameter to about the residual scale, as with huber_slope in XGBoost’s reg:pseudohubererror.

The legacy SnapBoost class also accepts mode ("classification" or "regression") and is deprecated. The legacy SnapBoost_KernelRidge class is also deprecated; it is the one exception to the table above and defaults to verbose=True. Invalid constructor values are rejected at fit, not at construction.

SnapBoost-specific

Parameter	Type	Default	Description
p_tree	float	0.9	Total probability of the tree families, split evenly across depths
p_linear	float	0.0	Probability allocated to the optional weighted linear family
min_max_depth	int	2	Minimum max_depth for trees in the pool
max_max_depth	int	4	Maximum max_depth for trees in the pool
min_samples_leaf	int	10	Minimum samples per leaf for decision trees
alpha	float	1.0	L2 regularization for the RFF ridge regressor
alpha_linear	float or None	None	L2 penalty for the optional linear family; defaults to alpha
gamma	float or "scale"	1.0	Kernel coefficient, or sklearn’s variance-based 'scale'
n_components	int	100	Number of random Fourier features
scale_features	bool	True	Standardize inputs used by the RFF learner
max_features	None, int, float, or str	None	Features considered by each tree split
kernel_gammas	sequence or None	None	Optional kernel bandwidth pool
kernel_types	sequence	("rbf",)	Any combination of "rbf" and "laplacian"
monotonic_cst	sequence or None	None	Per-feature monotonic tree constraints when supported; binary and regression only

Tip

On problems with both piecewise and smooth structure, try p_tree around 0.8–0.9. Setting p_tree=1.0 disables the ridge learner; p_tree=0.0 uses ridge only.

Label conventions

For binary and multiclass classification, original labels are accepted. Predictions use those labels and probability columns follow classes_ order. Binary decision_function is one-dimensional; multiclass returns one column per class.

Fit-time data

fit(X, y, sample_weight=None, eval_set=None, *, eval_sample_weight=None) accepts non-negative observation weights, one validation pair, and optional validation weights. early_stopping_rounds has an effect only when eval_set=(X_validation, y_validation) is provided. eval_metric(y, raw) receives the original labels, not the internal {-1, +1} encoding.

eval_metric and callbacks are optional fit-time arguments. candidate_n_jobs controls threaded candidate fitting in greedy mode only. After fit, staged_predict (and classifier staged_predict_proba / staged_decision_function) yield the ensemble after each round. permutation_importance(X, y) is the recommended feature-importance API for mixed tree and kernel learners.