def load_franke_data(self, cfg: CN, perm_index):
x, y, z = create_frankie_data(cfg.SEED, cfg.DATA.FRANKIE.N,
cfg.DATA.FRANKIE.NOISE)
X = create_X(x, y, cfg.DATA.FRANKIE.P)
self.split_and_scale_train_test(X,
z,
perm_index,
test_size=cfg.TEST_SIZE)
return self
def create_test_data():
seed = 3155
N = 20
noise_strength = 0.1
p = 5
x, y, z = create_frankie_data(seed, N, noise_strength)
X = create_X(x, y, p)
return X, z
# Cross valid and bootstrap for different alphas
N = 40
noise = 0.3
p = 4
alphas = [0.0, 0.000001, 0.00001, 0.0001, 0.001, 0.01, 0.1, 1.0, 10.0, 100.0]
l = len(alphas)
trials = 1000
sample_count = N
kfolds = 5
x, y, z = create_frankie_data(SEED, N, noise_strength=noise)
perm_index = np.random.permutation(len(z))
mse_boot = np.zeros(l)
mse_kfold = np.zeros(l)
mse_boot_train = np.zeros(l)
mse_kfold_train = np.zeros(l)
bias_boot = np.zeros(l)
var_boot = np.zeros(l)
for i in range(len(alphas)):
progressBar(i + 1, l)
X = create_X(x, y, p, debug=False)
crossval = CrossValidationKFold(kfolds).train_and_test(