Talk

functime: a next generation ML forecasting library powered by Polars

giovedì 23 maggio

11:00 - 11:30
StanzaTagliatelle
LinguaInglese
Livello audienceBeginner
Elevator pitch

Powered by this impressive query engine, functime enables forecasting thousands of time series at once, from the comfort of your laptop. With Polars, we can push the boundary for what “reasonable scale” means - and build a new generation of tools for machine learning.

Descrizione

Polars is mature, production ready, intuitive to write and pleasant to read. And it’s fast. Thanks to Rust and Rayon, you can achieve speeds greater than numba’s. If you combine it with top-of-the-class evaluation methods, not only can you get speedups of about 1-2x order of magnitude in feature engineering and cross-validation, but also dramatically improve your development workflow.

That’s what we set out to demonstrate with functime. We chose to write a time-series library first, because forecasting can be a costly undertaking, with significant problems of scale. Making predictions with big panel datasets usually required fitting thousands of univariate models, one at a time, using distributed systems. On the other hand, functime unlocks an efficient forecasting workflow, from your laptop.

📍Intended audience. This talk is a hands-on demonstration for forecasting practitioners and data scientists alike. It will showcase how to build clean and performant forecasting pipelines with rich feature-engineering capabilities - enabling a seamless and more efficient modelling workflow.

Nevertheless, the principles behind functime can be grasped by every machine learning practitioner: forecasting is just a use-case to shows off Polars’ potential. With Polars, we can improve the current state of machine learning modelling and raise the ceiling for what reasonable scales means.

🚩 Talk outline

• minutes 0-3. Problem setting: the current problem with forecasting. • minutes 3-7. What is Polars and why it is so fast. • minutes 7-10. What is global forecasting and why it is so effective. • minutes 10-20. A simple fit-evaluate modelling workflow. • minutes 20-25. An advanced workflow with blazingly fast feature extractors and cross-validation. • minutes 25-30. Wrap up and QA.

TagsBig Data, Machine-Learning, Databases, Data Structures
Participant

Luca Baggi

ML Engineer @xtream by day, and functime core developer by night, Luca is an absolute Polars fanboy - so much so he spoke about it at several Python Conferences, including the last PyData Global 🐻‍❄️ Feel free to reach out if you have any questions.