objectif

Tracer learning curve et validation curve pour diagnostiquer biais/variance.

code minimal

from sklearn.datasets import load_iris
from sklearn.model_selection import learning_curve, validation_curve
from sklearn.svm import SVC

X, y = load_iris(return_X_y=True)
train_sizes, train_scores, test_scores = learning_curve(SVC(), X, y, train_sizes=[0.5,0.8,1.0], cv=3)
param_range = [0.1, 1, 10]
vc_train, vc_test = validation_curve(SVC(), X, y, param_name="C", param_range=param_range, cv=3)
print(len(train_sizes), len(param_range))

utilisation

print(train_scores.shape[1] == 3)

variante(s) utile(s)

print(vc_test.shape[1] == 3)

notes

• Utiliser pour choisir n d’échantillons et hyperparamètres.