search

Nixtla / nixtla

TimeGPT-1: production ready pre-trained Time Series Foundation Model for forecasting and anomaly detection. Generative pretrained transformer for time series trained on over 100B data points. It's capable of accurately predicting various domains such as retail, electricity, finance, and IoT with just a few lines of code 🚀.
https://docs.nixtla.io
Other
1.91k stars 151 forks source link

[Feature Request]: Conformal CDFs #373

Open galenseilis opened 1 month ago

galenseilis commented 1 month ago

I'm excited by the prospect of uncertainty quantification via conformal prediction that has been implemented. I noticed that it can do quantiles and prediction intervals in the current state. Would it be possible to get conformal prediction cumulative distribution functions?

An example package that does this is crepes.

elephaint commented 1 month ago

At this point we don't have that implemented, but it might be something that we implement later on. For now, to create (i)cdf's, you could increase the grid size of the levels used when forecasting with TimeGPT (similar as to what crepes suggests in case of a large calibration set)