models.grid_search()

Scores every candidate by cross_validate on the same splits and

Usage

Source

models.grid_search(
    candidates,
    make,
    design,
    y,
    splits,
    score,
    n_jobs=None,
)

picks the lowest mean score.

Parameters

candidates: list

Hyperparameter values, in any form make accepts.

make: callable

make(candidate) -> model.

design: Design

y: list of float

splits: list of (list of int, list of int)

score: callable

As cross_validate.

n_jobs: int = None

Returns

SearchResult

Examples

>>> from prospicio.models import Design, ElasticNet, deviance_score, grid_search, k_fold
>>> x = [i / 10 for i in range(40)]
>>> d = Design([[1.0] * 40, x], ["(Intercept)", "x"])
>>> y = [1.0 + 2.0 * v for v in x]
>>> found = grid_search([0.0, 1.0, 10.0], lambda lam: ElasticNet("gaussian", lam=lam),
...                     d, y, k_fold(40, 4, 1), deviance_score("gaussian"))
>>> found.best_candidate

0.0