Mikio L. Braun

ML and AI expert

Back to archive

Thread

2 tweets

1
@rasbt Yeah, non-linear SVMs definitely. But I think linear SVMs and ridiculously large sparse manually constructed feature spaces were powering the world for a while. Then gradient boosted trees took over.
1
2
@paul_rietschka @rasbt Yeah, and don‘t get me started about the theoretical guarantees. Technically not wrong but I think there was practically quite a leap of faith. Those were worst case bounds that didn‘t explain while „average case“ it worked so well.
1