On: Why do tree-based models still outperform deep lea...

[redox99]

In what way have models gotten better for tabular data? Can't think of any new technique since 2022.

[jeffreyrogers → redox99]

There has been some work on training on lots of different data sets and then specializing on the one you care about. But I think people were trying that approach pre-2022 as well.

[asdff → jeffreyrogers]

This has to be done with great care. Most datasets are of poor quality.

[frituur → jeffreyrogers]

Do you have some good scientific references for that? I'd love to incorporate them in my phd thesis!

[jeffreyrogers → frituur]

Sorry, I don't have references off the top of my head. I just recall coming across it while I was working on something related to timeseries forecasting.

[_pastel → redox99]

Tooling around embeddings has improved. Creating and fine-tuning custom embeddings for your tabular data should be easier and more powerful these days.