Distance benchmarks

In this example, we compare the performance of dynamic time warping as implemented in wildboar against the implementation in sktime.

import warnings
warnings.filterwarnings("ignore")

import numpy as np
from wildboar.distance import pairwise_distance

from sktime.distances import ddtw_distance

rnd = np.random.RandomState()

x = rnd.randn(10000)
y = rnd.randn(10000)

%timeit pairwise_distance(x, y, metric="ddtw", metric_params={"r": 0.5})

304 ms ± 1.28 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)

%timeit ddtw_distance(x, y, window=0.5)

4.41 s ± 68.1 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)

As we can see, the wildboar implementation is more than one order of magnitude faster than the implementation in sktime.