Make kNN more resilient to nominal data

online-ml / river

🌊 Online machine learning in Python

https://riverml.xyz

BSD 3-Clause "New" or "Revised" License

5.03k stars 540 forks source link

Closed e10e3 closed 3 months ago

e10e3 commented 3 months ago

Fixes #1565 by using a different formula for the difference when the data is not numeric.

This allows the distance not to error out when non-numbers are given, while keeping a reasonable behaviour.

Because of the added code paths that are difficult to optimise away, kNN is roughly twice as slow to make a prediction.

smastelini commented 3 months ago

Please, see my comments on #1565.