HALHyperparameters

Properties

max_degree

integer | object≥ 1

Highest order of interaction among features that a basis function may represent. Cannot exceed the number of features in the training data.

smoothness_orders

integer | object≥ 0≤ 9

Smoothness of the basis functions: 0 fits zero-order indicators, 1 piecewise linear splines, higher values higher-order splines.

num_knots

array | null≥ 1 items

Number of knots per interaction degree, one value per degree, non-increasing. NULL generates them from max_degree and smoothness_orders.

reduce_basis

number | object | null≤ 1> 0

Minimum proportion of cases a basis function must be non-zero in to be kept. NULL uses the backend default of 1/sqrt(n). Applies only when smoothness_orders is 0.

max_basis

integer≥ 1

Largest projected basis-function count that will be fit. Training aborts above it rather than enumerating a basis that will not finish.

cv_select

boolean

Select lambda by cross-validation inside the fit; determined by the class.

use_min

boolean

Select lambda.min from the internal cross-validation. FALSE selects the more heavily penalized lambda.1se.

nfolds

integer≥ 3

Number of folds of the internal cross-validation that selects lambda.

seed

integer | null

Random seed for the internal cross-validation's fold assignment. NULL leaves it drawn from the ambient RNG.

ifw

boolean | object

Inverse Frequency Weighting in classification.

Relationships