tpe.py

from sklearn.datasets import load_iris
from sklearn.neighbors import KNeighborsClassifier
from sklearn.model_selection import cross_val_score

from hyperactive import Hyperactive
from hyperactive.optimizers import TreeStructuredParzenEstimators


data = load_iris()
X, y = data.data, data.target


def model(opt):
    knr = KNeighborsClassifier(n_neighbors=opt["n_neighbors"])
    scores = cross_val_score(knr, X, y, cv=5)
    score = scores.mean()

    return score


search_space = {
    "n_neighbors": list(range(1, 100)),
}

hyper = Hyperactive()
hyper.add_search(model, search_space, n_iter=100)
hyper.run()

search_data = hyper.search_data(model)


optimizer = TreeStructuredParzenEstimators(
    gamma_tpe=0.5, warm_start_smbo=search_data, rand_rest_p=0.05
)

hyper = Hyperactive()
hyper.add_search(model, search_space, optimizer=optimizer, n_iter=100)
hyper.run()