gtancev

# Import library.
import pymc3 as pm

# Import data.
filename = "wine.csv"
data = pd.read_csv(filename, sep=';', header=0, index_col=False)
data["Class"][data["Class"] > 1] = 0
data.head()

# Define model.

gtancev

# Define Gaussian kernel.
def Gaussian_kernel(x, x0, gamma):
	return (1/np.sqrt(2*np.pi*gamma)) * np.exp(-(x - x0)**2/(2*gamma))

gtancev

# Define kernel density.
def p(x, x0, gamma, kernel=Gaussian_kernel):
    p_x = 0
    for num in x0:
        p_x += kernel(x, num, gamma)
    return 1/(len(x0)) * p_x

gtancev

from sklearn.model_selection import KFold

folds = 10
kf = KFold(n_splits=folds)

gammas = np.logspace(-2, 1, 51)
log_proba = np.zeros((gammas.shape[0]))

# Perform cross-validation.
for i, gamma in enumerate(gammas):

gtancev

# Define y = f(x).
def f_y(x, x0, y0, gamma, kernel=Gaussian_kernel):
    assert len(x0) == len(y0)
    f_y = 0
    f_x = 0
    for k in range(len(x0)):
        f_y += kernel(x, x0[k], gamma)*y0[k]
        f_x += kernel(x, x0[k], gamma)
    return f_y / f_x

gtancev

folds = 10
kf = KFold(n_splits=folds)

gammas = np.logspace(-4, -1, 51)
mse = np.zeros((gammas.shape[0]))

# Perform cross-validation.
for i, gamma in enumerate(gammas):
    mse_ = 0
    for train_index, test_index in kf.split(x0):