Neural Prophet is compatible with the SoftDTWLossPytorch function from tslearn

[YannCabanes]

Hello,

I am a maintainer of the tslearn python package about time series learning.
We recently added the support of the PyTorch backend. In particular, we have added the class SoftDTWLossPytorch.
In the issue tslearn-team/tslearn#477 of tslearn, we have been asked if Neural Prophet can be used with SoftDTWLossPytorch.

We can combine the SoftDTWLossPytorch from tslearn with NeuralProphet from neuralprophet by defining:

from tslearn.metrics import SoftDTWLossPyTorch
from tslearn.metrics.soft_dtw_loss_pytorch import _SoftDTWLossPyTorch

def soft_dtw_loss_function(x, y, dist_func=SoftDTWLossPyTorch._euclidean_squared_dist, gamma=0.1):
    d_xy = dist_func(x, y)
    return _SoftDTWLossPyTorch.apply(d_xy, gamma)

and then:

m = NeuralProphet(loss_func=soft_dtw_loss_function)

[YannCabanes]

In the issue tslearn-team/tslearn#477, I propose a code which is running on Google Colab. This code is inspired by Neural Prophet tenth notebook (https://github.com/ourownstory/neural_prophet/blob/main/docs/source/tutorials/tutorial10.ipynb) and demonstrates the compatibility between NeuralProphet and the function SoftDTWLossPytorch from tslearn.