How to use to_time_series_dataset with a multidimensional dataset

[AndreasAZiegler]

Describe the bug
I have a dataset with 116 time series, every time series has 6 dimensions and a length of 578 elements. When I have the data the following way

len(dataset)
116
len(dataset[0])
6
len(dataset[0][0])
578

to_time_series_dataset(dataset) returns an array of shape (116, 6, 578) therefore (number of time series, dimensions, length of time series). However, the documentation mentions the data format (n_ts, max_sz, d).

To Reproduce
dataset = to_time_series_dataset(dataset)

Expected behavior
dataset.shape = (116, 578, 6)

Environment (please complete the following information):

• OS: Linux
• tslearn version: 0.1.29-1

[rtavenar]

Hi @AndreasAZiegler

to_time_series_dataset does not reorder dimensions, so if your array (or array-like) is of shape (116, 6, 578), you will have to transpose it before passing it to to_time_series_dataset.