A DILATE loss implementation that supports batches of timeseries for multi-step forecasting.
The implementation is done in PyTorch and uses a callable torch.nn.Module.
See example jupyter notebook for usage.
To install python package use git clone https://github.com/marcdemers/batch-DILATE, then cd batch-DILATE, followed by pip install .
Original paper can be found here: https://arxiv.org/abs/1909.09020
The (univariate timeseries) implementation here: https://github.com/vincent-leguen/DILATE