def get_dft_coefs(X, n_coefs):
"""
:param X: the training set numpy array shape:[n_samples, n_time_steps]
:param n_coefs: number of the coefficients
:return: numpy array shape: [n_samples, n_coefs]
"""
dft = DiscreteFourierTransform(n_coefs=n_coefs,
norm_mean=False,
norm_std=False)
X_dft = dft.fit_transform(X)
n_samples = len(X)
if n_coefs % 2 == 0:
real_idx = np.arange(1, n_coefs, 2)
imag_idx = np.arange(2, n_coefs, 2)
X_dft_new = np.c_[X_dft[:, :1],
X_dft[:, real_idx] + 1j * np.c_[X_dft[:, imag_idx],
np.zeros(
(n_samples, ))]]
else:
real_idx = np.arange(1, n_coefs, 2)
imag_idx = np.arange(2, n_coefs + 1, 2)
X_dft_new = np.c_[X_dft[:, :1],
X_dft[:, real_idx] + 1j * X_dft[:, imag_idx]]
return X_dft_new
import numpy as np
import matplotlib.pyplot as plt
from pyts.approximation import DiscreteFourierTransform
# Parameters
n_samples, n_timestamps = 100, 48
# Toy dataset
rng = np.random.RandomState(41)
X = rng.randn(n_samples, n_timestamps)
# DFT transformation
n_coefs = 30
dft = DiscreteFourierTransform(n_coefs=n_coefs,
norm_mean=False,
norm_std=False)
X_dft = dft.fit_transform(X)
# Compute the inverse transformation
if n_coefs % 2 == 0:
real_idx = np.arange(1, n_coefs, 2)
imag_idx = np.arange(2, n_coefs, 2)
X_dft_new = np.c_[X_dft[:, :1],
X_dft[:, real_idx] + 1j * np.c_[X_dft[:, imag_idx],
np.zeros((n_samples, ))]]
else:
real_idx = np.arange(1, n_coefs, 2)
imag_idx = np.arange(2, n_coefs + 1, 2)
X_dft_new = np.c_[X_dft[:, :1],
X_dft[:, real_idx] + 1j * X_dft[:, imag_idx]]
X_irfft = np.fft.irfft(X_dft_new, n_timestamps)
reconstructed5 = ssa.view_reconstruction(*[ssa.Xs[i] for i in streams5],
names=streams5,
return_df=True)
ssa.forecast_recurrent(steps_ahead=48, singular_values=streams10, plot=True)
ts_copy5 = ssa.ts.copy()
ts_copy5['Reconstruction'] = reconstructed5.Reconstruction.values
ts_copy5.plot(title='Original vs. Reconstructed Time Series')
plt.show()
# Functioning Discrete Fourer Transform Approach
from pyts.approximation import DiscreteFourierTransform
n_coefs = 21
dft = DiscreteFourierTransform(n_coefs=n_coefs,
norm_mean=False,
norm_std=False)
X_dft = dft.fit_transform(ticker.daily.adjclose.values.reshape(1, -1))
n_samples, n_timestamps = 1, ticker.daily.shape[0]
if n_coefs % 2 == 0:
real_idx = np.arange(1, n_coefs, 2)
imag_idx = np.arange(2, n_coefs, 2)
X_dft_new = np.c_[X_dft[:, :1],
X_dft[:, real_idx] + 1j * np.c_[X_dft[:, imag_idx],
np.zeros((n_samples, ))]]
else:
real_idx = np.arange(1, n_coefs, 2)
imag_idx = np.arange(2, n_coefs + 1, 2)
X_dft_new = np.c_[X_dft[:, :1],
X_dft[:, real_idx] + 1j * X_dft[:, imag_idx]]
X_irfft = np.fft.irfft(X_dft_new, n_timestamps)
def test_actual_results(params):
"""Test that the actual results are the expected ones."""
arr_actual = DiscreteFourierTransform(**params).fit_transform(X, y)
arr_desired = _compute_expected_results(X, y, **params)
np.testing.assert_array_equal(arr_actual, arr_desired)
def test_parameter_check(params, error, err_msg):
"""Test parameter validation."""
dft = DiscreteFourierTransform(**params)
with pytest.raises(error, match=re.escape(err_msg)):
dft.fit(X, y)
def test_fit_transform(params):
"""Test that fit and transform yield the same results as fit_transform."""
arr_1 = DiscreteFourierTransform(**params).fit(X, y).transform(X)
arr_2 = DiscreteFourierTransform(**params).fit_transform(X, y)
np.testing.assert_array_equal(arr_1, arr_2)