API reference

The full public API is seven names, all importable from the top-level package:

from chimeraboost import (
    ChimeraBoostRegressor,
    ChimeraBoostClassifier,
    ChimeraBoostQuantileRegressor,
    CustomObjective,
    metrics,
    quantile_metrics,
    warmup,
)

	What it is
ChimeraBoostRegressor	Gradient boosted oblivious trees for regression. Squared-error, absolute-error, quantile, Huber, and the log-link losses.
ChimeraBoostClassifier	Gradient boosted oblivious trees for classification. Binary and multiclass, with calibrated probabilities.
ChimeraBoostQuantileRegressor	A whole grid of conditional quantiles from one booster, with levels that cannot cross.
CustomObjective	Base class for writing your own regression loss.
metrics	Scoring a fitted regressor or classifier: error, skill, calibration. Behind model.report().
quantile_metrics	Scoring a predicted quantile grid: pinball loss, CRPS, coverage, interval score, PIT.
warmup	Pre-compile the numba kernels so the first fit or predict is not slow.

All three estimators are scikit-learn compatible: fit, then predict or predict_proba. For worked examples see Recipes, and for defaults and guidance on every option see Parameters.