The Quantile Trick

When you're doing a regression you're sometimes not so much interested in predicting the most "likely" value, sometimes you're more interested in predicting a spectrum of likely values. Put differently: you may be interested in predicting the quantiles of a distribution, instead of the median value. In this video we'll explain the quantile trick, which involves a pinball loss, to deal with these situations.
Video Chapters:
00:00 Drawing a dataset
00:58 Predicting Quantiles
03:04 Pinball Loss
07:06 Interactive Demo
09:45 Comparing Models
If you're interested in drawing data yourself, check out this project:
github.com/koaning/drawdata
The code for all of our videos can be found on this Github repository:
github.com/probabl-ai/youtube...
The code for this specific episode can be found here:
github.com/probabl-ai/youtube...
If you're keen to see more videos like this, you can follow us over at @probabl_ai.

[@Spinnen]

Just wanted to say I love these kind of videos - early days for the channel just wanted to give that positive feedback to continue

[@probabl_ai]

(Vincent here) Happy to hear it. Plenty more is on the way!

[@andreshoyosidrobo549]

Great video! I think you nailed the explanation. It is nice to see how to use Jupyter widgets as tools to explain ML-related concepts.

[@probabl_ai]

Thanks!
Speaking of widgets. Seen these?
th-cam.com/video/STPv0jSAQEk/w-d-xo.html
th-cam.com/video/goaBFxGhp6Y/w-d-xo.html

[@hernanebraga23]

Excellent video, I really like how you explain complex concepts and techniques

[@GarveRagnara]

But what about conformal predictions? :D

[@probabl_ai]

(Vincent here) Noted! There's a long todo list for ideas, but I agree conformal predictions deserve attention.

[@jdt12880]

How does the w parameter map to tge quantile?

[@alexmolasmartin]

how do you go from the free parameter of the pinball loss to the quantile you want to predict?

[@probabl_ai]

(Vincent here) I might recommend playing with the notebook to unravel this one. But I'll keep this idea in mind because it might be an interesting appendix short later.

[@nicolasr8710]

It would be interesting to compare to Bayesian linear regression.