Why tree gradients give you a boost

Gradient boosted trees are a powerful machine learning technique that appear a lot in practice. There's a lot to like about their performance but there are plenty of other details under the hood that are worth having a deeper look at. That's why we're doing a long series of videos about these models, start with this first one that's all about the intuition.
If you're curious about the code, you can find the notebook for this series here:
github.com/probabl-ai/youtube...
00:00 Introduction
00:57 Regression task
04:06 Code
05:55 Charts
09:25 Properties

[@luca_dev]

Did you change something about the audio? It sounds better than usual!

[@NuclearStr1der]

Fantastic and clear presentation as always, Vincent!

[@apachaves]

Awesome intuitive explanations! As always from Vincent.

[@cairoliu5076]

great work!

[@anaveenan]

Is there intuition on how boosting impact calibration of output probability?

[@probabl_ai]

(Vincent here) This might depend a lot on the hyperparameters of the full pipeline. In general though calibration is always a good thing to consider once a boosted pipeline is trained.