K-Means Clustering Time Series Data of Different Length

I have been trying to understand where the "mean" or centroid of each cluster is located after using the k-means algorithm with DTW for time series data of different length.

If you use k-medoids, it is clear where the representative time serie of each cluster is since it is a real existing time serie in the data. But how about in case of k-means?

Key points:

- time series data of different length

- no euclidean distance, but DTW

- where is the location of centroids estimated by k-means? -> question

Thanks in advance.

[Aditya Patil]

Can you share the code that you are using?