Learned about gradient descent in my ML class, didn't think it would be applicable to graphics too
keenan
@Sybil For sure. Gradient descent, and optimization more generally, is a tool with extremely broad applications---and a very old history that predates computer science. If you really want to give yourself to tackle all sorts of future challenges, I'd strongly recommend learning the fundamentals of optimization outside of an application context. Learning about optimization only through ML or graphics gives you an incomplete picture, and can provide some mixed-up ideas that don't translate well to other problems. For instance: terms like "learning rate" and "loss function" aren't used outside of ML; they're jargon that has become commonplace only in the past 10 years or so. And they make it hard to appreciate ideas like line search, which are often eschewed in ML for application-centric reasons, yet are extremely important for understanding broader optimization.
Learned about gradient descent in my ML class, didn't think it would be applicable to graphics too
@Sybil For sure. Gradient descent, and optimization more generally, is a tool with extremely broad applications---and a very old history that predates computer science. If you really want to give yourself to tackle all sorts of future challenges, I'd strongly recommend learning the fundamentals of optimization outside of an application context. Learning about optimization only through ML or graphics gives you an incomplete picture, and can provide some mixed-up ideas that don't translate well to other problems. For instance: terms like "learning rate" and "loss function" aren't used outside of ML; they're jargon that has become commonplace only in the past 10 years or so. And they make it hard to appreciate ideas like line search, which are often eschewed in ML for application-centric reasons, yet are extremely important for understanding broader optimization.