Project 20: Exploring Explainability for Time Series Representations Learned through Self-Supervised Learning.
In this project we explore the use of the RELAX method to provide explainability for representations of spectograms learned through both self-supervised and supervised learning.
Furthermore, we explore the use of segmentation based methods such as SINEX and a SINEXC inspired algorithm.
We test the methods using two different pre-trained models, therefore the repository is split into a folder for each model each containing an Explainer_notebook that reproduces the results from the report.