>>13934330>>13934344Why are you surprised? If anything, regression always beats out everyone if you have an underlying model (a.k.a. nonlinear regression btfo everyone else).
>>13939140It's not touted, it's like a couple classes in any given statistical degree. You can get a degree in statistics while barely touching. Some fields don't use it, in a lot of statistical modeling (like time series analysis) it's pretty well known that neural nets don't buy you much (if anything).
It's just corporate memes honestly. I'm not shitting on neural nets either, they're a beautiful method honestly for piecing together nonlinear relationships, but like, if there's not already a clear relationship that a human can see, you probably aren't going to buy too much with a neural net.
Anyway, the cool parts of 'learning' are in the unsupervised community (and quite a few techniques here don't even involve neural nodes like most clustering algorithms) anyway.
>>13939190They don't buy you anything and even basic ARIMA models (which is literally just linear regression fixed for time series data) can beat them or perform just as well for no cost and easy interpretability.
Again, the real question is, why are you surprised? If anything, I'm surprised when things BEAT out linear regression. Regression is elegant as fuck.
