News for hackers. Stuff that matters.

A comment on The Matrix Calculus You Need for Deep Learning. Parent comment

Re:The Matrix Calculus You Need for Deep Learning by thomasahle on Friday June 29, 2018 @04:54AM (#17423285)
> I'm not sure how you saw [the Hessian] used for fitting a logistic model

Probably some 2nd derivative version of gradient descent or Newton's method.

https://en.wikipedia.org/wiki/Newton%27s_method_in_optimizat...

[ Parent | on HN ]
Re:The Matrix Calculus You Need for Deep Learning by mehrdadn on Friday June 29, 2018 @06:19AM (#17423615)
Yeah that's what I was referring to when I said "approximating it as a 2nd-order polynomial and minimizing that".
[ Parent | on HN ]