CODEWITHSUNDEEP
pythontime-seriesdtw
eva
eva

Reputation: 1

different distance depending on input order using fastdtw in python

I want to find similarities between time-series and perform Dynamic Time Warping in python using fastdtw (https://github.com/slaypni/fastdtw/tree/master/fastdtw). The computed distance is sometimes changing according to the input order, especially for long time-series. Here is an example:

x = np.random.rand(44)
y = np.random.rand(40)
d1, inx1 = fastdtw(y, x)
d2, inx2= fastdtw(y, x)
d1, d2

Any idea why?

Upvotes: 0

Views: 1317

Answers (1)

eamonn
eamonn

Reputation: 101

Why use fastDTW which is approximate and (in spite of its name, slow), when you can compute DTW fast and exact? https://www.cs.unm.edu/~mueen/DTW.pdf

Upvotes: 1

Related Questions