Hacker News new | past | comments | ask | show | jobs | submit login

Differential equations and electromagnetism?

Maybe 30 years ago, but these days electrical engineers know how to code

There seems to be a tendency to exaggerate the breadth of things that people should study. When you ask for specific instances where something would be useful, you usually get a hand wavy answer that since X field is connected to what you want to study, it is therefore worthwhile to study X field ("Of course computer scientists should study advanced chemistry, we wouldn't even have computers without chemistry! And what if you needed to written a program for chemists!").

In my experience people retain very little of something if they don't keep using it. Many people take things like calculus in highschool or college, and completely forget it within a few years. I often checked with classmates about how much they retained from courses we had just taken the previous semester, and most of the stuff that wasn't directly connected to what they were currently studying seemed to be forgotten (if they ever had a real understanding of it to begin with).

So pair calculus with a programming project that involves physics simulation.

Who is going to write FEA software? Or rigid body dynamics simulation? Or the physically based rendering software? It's not going to be a typical electrical engineer.

Isn't modern ML based on differential equations?

Gradient descent is partial differentiation, and Lagrange multipliers come up in the context of nearest-neighbor searches. I’m almost positive most people just use it without actually understanding the theory behind why it works, though.

Sure, but only a handful of people really employ advanced calculus for building ML, while the vast majority of Data Scientist only use the ML algorithms implementations as a black box, without diving into its inner workings.

I realize that there is a certain conflict here, since I'm happy to program computers without understanding the physics of transistors, but your comment makes me nervous.

I imagine the DE part of ML is, practically speaking, imported from a library somewhere by practitioners.

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact