Patricio Cerda-Mardini

From an ML point of view it's not immediately clear how to tackle this, but I suspect defining a distribution over values instead of a point sample may be the...

I think we could replace the "Predictor-specific methods" section with one called "ML Handlers" or similar, and inside simply link to the new page added in #4037, explaining that their...

Yes, the horizon parameter definitely has an impact on the training procedure if using the Lightwood ML backend. It can be a notorious change, for example if `horizon=1` the mixers...

Is it used in Ludwig? As far as I know that dep comes from here: https://github.com/mindsdb/mindsdb/pull/3225/files#r992412981 EDIT: it is quite weird though how the version installed (6.0.1) is 1) older...

Hi @MinuraPunchihewa! I've got an update: we’ve simplified the pipeline so that data analysis is now its own package (at [`mindsdb/dataprep_ml`](https://github.com/mindsdb/dataprep_ml)), among other changes. Thanks to this, while `lightwood` still...

Is metabase connected to a local installation of MindsDB? Or cloud?

Also: add docs within Lightwood that reference the evaluator package, and what functions are supported.

LightGBM is no longer a default option, so it's not happening as default behavior. However, if installed and/or user opts for another boosting or tree-based approach (e.g. XGBoost) then it...

Seems important for user experience, so I would say yes.

@parthiv11 static code check is failing, try `flake8 ./` in your handler folder to spot all the relevant lines that must be fixed