0:00
/
Generate transcript
A transcript unlocks clips, previews, and editing.

Conformal Seasonal Pools: Efficient Training-Free Probabilistic Time-Series Forecasting

Delighted to advise that Conformal Seasonal Pools has been accepted to the COPA 2026 conference.

A simple model that requires no training outperforms Amazon’s DeepNTPS whilst beein x500 faster.

https://arxiv.org/abs/2605.03789

https://github.com/valeman/csp-forecaster

Discussion about this video

User's avatar

Ready for more?