skglm.experimental.Pinball

class skglm.experimental.Pinball(quantile_level)

Pinball datafit.

The datafit reads:

sum_i quantile_level * max(y_i - Xw_i, 0) +
(1 - quantile_level) * max(Xw_i - y_i, 0)

with quantile_level in [0, 1].

Parameters:

quantile_levelfloat

Quantile level must be in [0, 1]. When quantile_level=0.5, the datafit becomes a Least Absolute Deviation (LAD) datafit.

__init__(quantile_level)

Methods

__init__(quantile_level)
get_spec()	Specify the numba types of the class attributes.
initialize(X, y)	Pre-computations before fitting on X and y.
initialize_sparse(X_data, X_indptr, X_indices, y)	Pre-computations before fitting on X and y when X is a sparse matrix.
params_to_dict()	Get the parameters to initialize an instance of the class.
prox(w, step, y)	Prox of step * pinball.
prox_conjugate(z, step, y)	Prox of step * pinball^*.
subdiff_distance(Xw, z, y)	Distance of z to subdiff of pinball at Xw.
value(y, w, Xw)	Value of datafit at vector w.