>>12931057You've been posting this in the /mcs/ thread and now here. Theoretical ML isn't dead, but it's definitely not nearly as popular as empirical work. I doubt you'd see a theorist like, say, Arora push out garbage. But there definitely needs to be a push for higher integrity among ML researchers - ie, squash the rat race.
>>12931164>fancy word for statsThere was a good prof from Stanford who had a statement about this. Statistics definitely has a huge part in ML, but we can't reduce all of ML down to stats since the other half of the machinery we use is the research on massive data retrieval. The computational side, both in the implementation and learnability, is pretty important too.
But the further you get into ML theory, the more you lose any distinction between "this is statistics" and "this is analysis" and "this is computational theory." It just bleeds together into ML. You don't end up caring whether classifying the popular Hilbert spaces is more functional analysis or machine learning.
