https://github.com/google-research/soft-dtw-divergences

An implementation of soft-DTW divergences.
https://github.com/google-research/soft-dtw-divergences

dtw dynamic-time-warping soft-dtw time-series

Last synced: about 2 months ago
JSON representation

An implementation of soft-DTW divergences.

• Host: GitHub
• URL: https://github.com/google-research/soft-dtw-divergences
• Owner: google-research
• License: apache-2.0
• Created: 2020-11-18T22:05:43.000Z (almost 4 years ago)
• Default Branch: master
• Last Pushed: 2021-10-14T14:27:20.000Z (almost 3 years ago)
• Last Synced: 2024-06-17T13:35:27.980Z (3 months ago)
• Topics: dtw, dynamic-time-warping, soft-dtw, time-series
• Language: Python
• Homepage:
• Size: 16.6 KB
• Stars: 131
• Watchers: 9
• Forks: 16
• Open Issues: 1
• Metadata Files:
  • Readme: README.md
  • Contributing: CONTRIBUTING.md
  • License: LICENSE

Awesome Lists containing this project

README

        
Differentiable Divergences between Time Series

==============================================

An implementation of soft-DTW divergences.

Example

-------

```python

import numpy as np

from sdtw_div.numba_ops import sdtw_div, sdtw_div_value_and_grad

# Two 3-dimensional time series of lengths 5 and 4, respectively.

X = np.random.randn(5, 3)

Y = np.random.randn(4, 3)

# Compute the divergence value. The parameter gamma controls the regularization strength. 

value = sdtw_div(X, Y, gamma=1.0)

# Compute the divergence value and the gradient w.r.t. X.

value, grad = sdtw_div_value_and_grad(X, Y, gamma=1.0)

```

Similarly, we can use `sharp_sdtw_div`, `sharp_sdtw_div_value_and_grad`,

`mean_cost_div` and `mean_cost_div_value_and_grad`.

Install

--------

Run `python setup.py install` or copy the files to your project.

Reference

----------

> Differentiable Divergences between Time Series 


> Mathieu Blondel, Arthur Mensch, Jean-Philippe Vert 


> [arXiv:2010.08354](https://arxiv.org/abs/2010.08354)