ML-Baselines
https://ludwigstumpp.com/ml-baselines
During my two years as an AI Engineer @appliedAI I learned to appreciate the iterative approach to designing and testing machine learning models by using simple baselines.
A baseline can help to:
- Set a lower bound on your expected final model performance. This is also great to communicate with stakeholders.
- Check that everything apart the model is working (Dataloader, Deployment, ...).
- Helps to prevent future errors as more advanced models are expected to achieve greater performance. If not, something is probably off.
- Allow you to play around and try out a few metrics / losses.
- Allow a possible client to try out the planned model interface (API, Dashboard)
To help a little bit with this engineering paradigm, I created ML-Baselines, a collection of baselines for typical machine learning problems and tools that help you with it.
Have fun 🤗.
Have fun 🤗.