Reproducing "ClaSP - Time Series Segmentation"
Siegfried Schwartz, Lenz, Philipp, Leonard Galustian · Zenodo (CERN European Organization for Nuclear Research) · 2023
In the paper "ClaSP - Time Series Segmentation", a novel unsupervised learning approach for Time Series Segmentation is introduced. In this work, we tried to reproduce the main results of this paper and therefore determine its reproducibility. Although some of the code was given and the description of the setup seemed quite clear at first, we did encounter severe obstacles including numerical instabilities, and missing/incomplete code with a number of bugs ultimately limiting the reproducibility of the experiments.