Hello everyone,
I’ve started working on a small early prototype related to my proposed GSoC 2026 idea on failure analysis and debugging for machine learning models on time-series data.
At the moment, the repository focuses on a very simple overfitting detection utility, along with a runnable example and clear instructions. The goal was to first set up a clean structure and make sure the project is easy to run and understand before adding more advanced functionality.
Repository:
I would really appreciate any feedback on the direction, structure, or suggestions for what failure modes or diagnostics might be most useful to prioritize next.
Thanks in advance!