Multidimensional/multivariate dynamic time warping (DTW) library/code in Python

Question

I am working on a time series data. The data available is multi-variate. So for every instance of time there are three data points available. Format:

|

Answer 1

I think that it is a good idea to try out a method in whatever implementation is already available before considering whether it worth working on a reimplementation.

Did you try the following ?

from rpy2.robjects.packages import importr
# You'll obviously need the R package "dtw" installed with your R
dtw = importr("dtw")

# all functions and objects in the R package "dtw" are now available
# with `dtw.<function or object>`

Answer 2

Thanks @lgautier I dug deeper and found implementation of multivariate DTW using rpy2 in Python. Just passing the template and query as 2D matrices (matrices as in R) would allow rpy2 dtw package to do a multivariate DTW. Also if you have R installed, loading the R dtw library and "?dtw" would give access to the library's documentation and different functionalities available with the library.

For future reference to other users with similar questions: Official documentation of R dtw package: https://cran.r-project.org/web/packages/dtw/dtw.pdf Sample code, passing two 2-D matrices for multivariate DTW, the open_begin and open_end arguments enable subsequence matching:

import numpy as np
import rpy2.robjects.numpy2ri
rpy2.robjects.numpy2ri.activate()
from rpy2.robjects.packages import importr
import rpy2.robjects as robj

R = rpy2.robjects.r
DTW = importr('dtw')

# Generate our data
template = np.array([[1,2,3,4,5],[1,2,3,4,5]]).transpose()
rt,ct = template.shape
query = np.array([[1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16],[1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16]]).transpose()
rq,cq = query.shape

#converting numpy matrices to R matrices
templateR=R.matrix(template,nrow=rt,ncol=ct)
queryR=R.matrix(query,nrow=rq,ncol=cq)

# Calculate the alignment vector and corresponding distance
alignment = R.dtw(templateR,queryR,keep=True, step_pattern=R.rabinerJuangStepPattern(4,"c"),open_begin=True,open_end=True)

dist = alignment.rx('distance')[0][0]

print dist

Answer 3

It seems like tslearn's dtw_path() is exactly what you are looking for. to quote the docs linked before:

Compute Dynamic Time Warping (DTW) similarity measure between (possibly multidimensional) time series and return both the path and the similarity.

[...]

It is not required that both time series share the same size, but they must be the same dimension. [...]

The implementation they provide follows:

H. Sakoe, S. Chiba, “Dynamic programming algorithm optimization for spoken word recognition,” IEEE Transactions on Acoustics, Speech and Signal Processing, vol. 26(1), pp. 43–49, 1978.