Example #1
0
def test_window_slice_windows():
    windows = SlidingWindow(size=4, stride=2)
    X = signal_embedded_search
    X_windows = windows.fit_transform(X)
    slice_idx = windows.slice_windows(X)
    assert_almost_equal(
        np.stack([X[begin:end] for begin, end in slice_idx]), X_windows
        )
Example #2
0
def test_window_resample():
    windows = SlidingWindow(width=3, stride=2)
    windows.fit(y)
    y_resampled = windows.resample(y)
    assert_almost_equal(y_resampled, y[np.arange(3, 20, 2)])
Example #3
0
def test_window_transform():
    windows = SlidingWindow(width=3, stride=2)
    x_windows = windows.fit_transform(signal_embedded_search)
    assert (x_windows.shape == (8, 4, 2))
Example #4
0
def test_window_params():
    windows = SlidingWindow(width=0)
    with pytest.raises(ValueError):
        windows.fit(signal)
Example #5
0
n_timestamps = 10
X, y = np.arange(n_timestamps), np.arange(n_timestamps) - n_timestamps
X, y

# We can instantiate our sliding window transformer-resampler and run it on the pair ``(X, y)``:

# 我们可以实例化滑动窗口转换器-重采样器，然后在（X，y）对上运行它

# In[2]:

from gtda.time_series import SlidingWindow

window_size = 3
stride = 2

SW = SlidingWindow(size=window_size, stride=stride)
X_sw, yr = SW.fit_transform_resample(X, y)
X_sw, yr

# We note a couple of things:
# - ``fit_transform_resample`` returns a pair: the window-transformed ``X`` and the resampled and aligned ``y``.
# - ``SlidingWindow`` has made a choice for us on how to resample ``y`` and line it up with the windows from ``X``: a window on ``X`` corresponds to the *last* value in a corresponding window over ``y``. This is common in time series forecasting where, for example, ``y`` could be a shift of ``X`` by one timestamp.
# - Some of the initial values of ``X`` may not be found in ``X_sw``. This is because ``SlidingWindow`` only ensures the *last* value is represented in the last window, regardless of the stride.
#
# 注意以下几点：
# fit_transform_resample返回一对：经过窗口转换的X和经过重新采样并对齐的y。
# SlidingWindow已为我们选择了如何对y重新采样并将其与X中的窗口对齐：X上的窗口与y上相应窗口中的最后一个值相对应。这在时间序列预测中很常见，例如，y可能是X偏移一个时间戳。
# X的某些初始值可能在X_sw中找不到。这是因为SlidingWindow仅确保最后一个值在最后一个窗口中表示，而不管步幅如何。

# ## Multivariate time series example: Sliding window + topology ``Pipeline``
Example #6
0
def test_window_plot():
    windows = SlidingWindow(width=3, stride=2)
    X_windows = windows.fit_transform(signal_embedded_search)
    windows.plot(X_windows, sample=0)
Example #7
0
def test_window_params(size):
    windows = SlidingWindow(size=size)
    with pytest.raises(ValueError):
        windows.fit(signal)
Example #8
0
def test_window_transform():
    windows = SlidingWindow(size=4, stride=2)
    X_windows = windows.fit_transform(signal_embedded_search)
    assert (X_windows.shape == (8, 4, 2))