Implementation of soft dynamic time warping in pytorch
Here is my implementation of the Soft Dynamic Time Warping loss function described in https://arxiv.org/abs/1703.01541.
Currently I have only a 'naive' implementation without extending the fast cython implementation in https://github.com/mblondel/soft-dtw to incorporate a batch dimension. If I continue to use this in my line of research I may implement a cython / CUDA version to increase speed.