>>13975559Hmm? Are they polynomial splines?
Anyway, yeah no shit. Anyone with half a brain and computational knowledge would realize this. Like polynomial splines already fit to data ridiculously well.
I mean, it makes sense too, what do you think you do in a single node? You do a regression, then in the next layer you do another regression of the regressions. What do you think a regression of regressions looks like?
You essentially, at each layer, are creating a linear combination of inputs that are combined into some basis functions and then those are combined further into a linear combination and so on.
The real absurdity is that anyone would think that it isn't regression. I blame corporate America and academia meming for money.
Anyway, many of the other methods such as svms are also regression and actually have direct relationships with techniques like logistic REGRESSION. This is all to be expected. Our understanding of lines and polynomials is immense and computationally they are the most efficient and simple mathematical structures we have.
