Mikio L. Braun

ML and AI expert

Back to archive

Thread

2 tweets

1
With everyone is focussing on LLMs, I'd like to remind people that "classical" methods like xgboost are as relevant as ever even for text based use cases (e.g. sentiment classification, document classification) because they use a fraction of the resources that an LLM does.
4 16
2
Sure, you can do document classification with an LLM and it'll probably work for much broader documents, but running an LLM requires a machine with a lot of GPU compute power, while you can run a xgboost model on standard hardware, for the fraction of the time.
5