def test_fast_dtw():
"""Testing 'fast_dtw'"""
# Parameter
x1 = np.array([1, -1, 1, 1, 1, -1])
x2 = np.ones(6)
x3 = -np.ones(6)
# Test 1
cost_actual = fast_dtw(x1, x2, window_size=2)
cost_desired = 2
np.testing.assert_equal(cost_actual, cost_desired)
# Test 2
cost_actual = fast_dtw(x1, x3, window_size=2)
cost_desired = 4
np.testing.assert_equal(cost_actual, cost_desired)
# Test 3
cost_actual = fast_dtw(x2, x3, window_size=2)
cost_desired = 6
np.testing.assert_equal(cost_actual, cost_desired)
def plot_fastdtw(x, y, window_size, dist='absolute', output_file=None):
"""Plot the optimal warping path between two time series subject to
a constraint region.
Parameters
----------
x : np.array, shape = [n_features]
first time series
y : np.array, shape = [n_features]
first time series
window_size : int
size of the window for PAA
dist : str or callable (default = 'absolute')
cost distance between two real numbers. Possible values:
- 'absolute' : absolute value of the difference
- 'square' : square of the difference
- callable : first two parameters must be real numbers
and it must return a real number.
output_file : str or None (default = None)
if str, save the figure.
"""
# Check input data
if not (isinstance(x, np.ndarray) and x.ndim == 1):
raise ValueError("'x' must be a 1-dimensional np.ndarray.")
if not (isinstance(y, np.ndarray) and y.ndim == 1):
raise ValueError("'y' must be a 1-dimensional np.ndarray.")
if x.size != y.size:
raise ValueError("'x' and 'y' must have the same size.")
# Size of x
x_size = x.size
# Check parameters
if not isinstance(window_size, int):
raise TypeError("'window_size' must be an integer.")
if window_size < 1:
raise ValueError("'window_size' must be greater or equal than 1.")
if window_size > x_size:
raise ValueError(
"'window_size' must be lower or equal than the size 'x'.")
if not (callable(dist) or dist in ['absolute', 'square']):
raise ValueError(
"'dist' must be a callable or 'absolute' or 'square'.")
region, D, path = fast_dtw(x,
y,
window_size=window_size,
approximation=False,
dist=dist,
return_path=True)
x_1 = np.arange(x_size + 1)
z_1 = np.zeros([x_size + 1, x_size + 1])
for i in range(x_size):
for j in region[i]:
z_1[j, i] = 0.5
for i in range(len(path)):
z_1[path[i][0], path[i][1]] = 1
plt.pcolor(x_1, x_1, z_1, edgecolors='k', cmap='Greys')
plt.xlabel('x', fontsize=20)
plt.ylabel('y', fontsize=20)
if output_file is not None:
plt.savefig(output_file)
# Show plot
plt.show()
import numpy as np
import matplotlib.pyplot as plt
from pyts.utils import fast_dtw
# Parameters
n_samples, n_features = 2, 48
# Toy dataset
rng = np.random.RandomState(41)
x, y = rng.randn(n_samples, n_features)
# Dynamic Time Warping
region, D, path = fast_dtw(x,
y,
dist='absolute',
window_size=6,
approximation=False,
return_path=True)
# Visualize the result
timestamps = np.arange(n_features + 1)
matrix = np.zeros([n_features + 1, n_features + 1])
for i in range(n_features):
for j in region[i]:
matrix[j, i] = 0.5
for i in range(len(path)):
matrix[path[i][0], path[i][1]] = 1
plt.figure(figsize=(8, 8))
plt.pcolor(timestamps, timestamps, matrix, edgecolors='k', cmap='Greys')
plt.xlabel('x', fontsize=20)