def test_actual_results_quantile_transformer_normal():
"""Test that the actual results are the expected ones."""
X_ppf = norm.ppf(np.linspace(0, 1, 1000)[1:-1])
weights = np.round(norm.pdf(X_ppf) * 1000).astype('int64')
X = []
for value, weight in zip(X_ppf, weights):
X += [value] * weight
X = np.asarray(X).reshape(1, -1)
transformer = QuantileTransformer(n_quantiles=11,
output_distribution='normal')
arr_actual = transformer.fit_transform(X)
arr_desired = X
atol = 0.01 * X.shape[1]
np.testing.assert_allclose(arr_actual, arr_desired, atol=atol, rtol=0.)
"""
import numpy as np
import matplotlib.pyplot as plt
from pyts.preprocessing import PowerTransformer, QuantileTransformer
# Parameters
n_samples, n_timestamps = 100, 48
# Toy dataset
rng = np.random.RandomState(41)
X = rng.randn(n_samples, n_timestamps)
# Scale the data with different scaling algorithms
X_power = PowerTransformer().transform(X)
X_quantile = QuantileTransformer().transform(X)
# Show the results for the first time series
plt.figure(figsize=(16, 6))
ax1 = plt.subplot(121)
ax1.plot(X[0], 'o-', label='Original')
ax1.set_title('Original time series')
ax1.legend(loc='best')
ax2 = plt.subplot(122)
ax2.plot(X_power[0], 'o--', color='C1', label='PowerTransformer')
ax2.plot(X_quantile[0], 'o--', color='C2', label='QuantileTransformer')
ax2.set_title('Transformed time series')
ax2.legend(loc='best')
def test_actual_results_quantile_transformer_uniform():
"""Test that the actual results are the expected ones."""
transformer = QuantileTransformer(n_quantiles=11)
arr_actual = transformer.fit_transform(X)
arr_desired = [np.linspace(0, 1, 11) for _ in range(3)]
np.testing.assert_allclose(arr_actual, arr_desired, atol=1e-5, rtol=0.)
* :class:`pyts.preprocessing.PowerTransformer`
* :class:`pyts.preprocessing.QuantileTransformer`.
This example illustrates the transformation from both algorithms.
"""
# Author: Johann Faouzi <*****@*****.**>
# License: BSD-3-Clause
import matplotlib.pyplot as plt
from pyts.datasets import load_gunpoint
from pyts.preprocessing import PowerTransformer, QuantileTransformer
X, _, _, _ = load_gunpoint(return_X_y=True)
n_timestamps = X.shape[1]
# Transform the data with different transformation algorithms
X_power = PowerTransformer().transform(X)
X_quantile = QuantileTransformer(n_quantiles=n_timestamps).transform(X)
# Show the results for the first time series
plt.figure(figsize=(6, 4))
plt.plot(X[0], '--', label='Original')
plt.plot(X_power[0], '--', label='PowerTransformer')
plt.plot(X_quantile[0], '--', label='QuantileTransformer')
plt.legend(loc='best', fontsize=8)
plt.title('Non-linear transformations', fontsize=16)
plt.tight_layout()
plt.show()