skglm.datafits.Huber

class skglm.datafits.Huber(delta)

Huber datafit.

The datafit reads:

${\displaystyle \frac{1}{n_{\text{samples}}}\sum _{i=1}^{n_{\text{samples}}}f(y_i-{\left(Xw\right)}_i)}$

where ${\displaystyle f}$ is the Huber function:

${\displaystyle f\left(x\right)=\{\begin{array}{ll}\frac{1}{2}{x}^2& ifx\le \delta \\ \delta \left|x\right|-\frac{1}{2}{\delta }^2& ifx>\delta \end{array}}$

Attributes:

deltafloat

Threshold hyperparameter.

__init__(delta)

Methods

__init__(delta)
full_grad_sparse(X_data, X_indptr, ...)
get_global_lipschitz(X, y)
get_global_lipschitz_sparse(X_data, ...)
get_lipschitz(X, y)
get_lipschitz_sparse(X_data, X_indptr, ...)
get_spec()	Specify the numba types of the class attributes.
gradient_scalar(X, y, w, Xw, j)
gradient_scalar_sparse(X_data, X_indptr, ...)
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.
intercept_update_step(y, Xw)
params_to_dict()	Get the parameters to initialize an instance of the class.
value(y, w, Xw)	Value of datafit at vector w.