This repository provides the code for the implementing Learning Sinkhorn Divergences for Change Point Detection.
Available change points are used to obtain triplet pairs which are then used to learn a metric for Sinkhorn divergences. This metric is then used by Sinkhorn divergences in two-sample tests over sliding windows for improved change point detection performance.
Requires
- Pytorch (1.8+)
- Python (3.7+)
- Matplotlib
- Scipy
Main file to run:
main_run.py