AdditiveChi2Sample (Additive Chi Squared Kernel)

Kernel:

k(x,y) = \sum_i \frac{2 x_i y_i}{x_i + y_i}

• Permite el muestreo de un kernel chi2 aditivo.

• Realiza un muestreo de la transformada de Fourier del kernel a intervalos regulares.

• Cada entrada en el espacio original es transformada en 2 * sample_steps- 1 características.

from sklearn.datasets import load_digits
from sklearn.kernel_approximation import AdditiveChi2Sampler
from sklearn.linear_model import SGDClassifier

X, y = load_digits(return_X_y=True)

chi2sampler = AdditiveChi2Sampler(
    # -------------------------------------------------------------------------
    # Gives the number of (complex) sampling points. 2n+1 points for dimension.
    # Typical values = {1, 2, 3}
    sample_steps=2,
    # -------------------------------------------------------------------------
    # Sampling interval. Must be specified when sample_steps not in {1,2,3}.
    sample_interval=None,
)

X_transformed = chi2sampler.fit_transform(X, y)

X_transformed.shape

(1797, 192)

clf = SGDClassifier(
    max_iter=100,
    random_state=0,
    tol=1e-3,
)

clf.fit(X_transformed, y)

clf.score(X_transformed, y)

0.9821925431274346