Theoretic vs Applied Math courses for gaining model intuition
retained text
Hi, I want to be clear, I have no intention of going into AI "research". I guess the distinction I could make is I want to know \*why\* a graph neural network works, vs coding it in pytorch and nuking it with more data to produce a better result (what I can do now), vs inventing a new algorithm mathematically from scratch (what I DO NOT want to do). Another example could be instead of researching online what fine-tuning parameters are beneficial, mathematically I could know what everything means and select the right one. But I do not want to invent a new fine-tuning parameter - I simply want to \*actually\* understand the existing ones. I have the choice to do more theoretic proof-based courses, or more applied ones. This means applied lin alg 2 vs proof-based linear algebra 2. I have done Calc 1 -> 3, Stats 1->2, Prob 1 -> 2, Discrete Math 1->2. I will be in this field for a long time, but again, I do not want to go into theoretical research whatsoever - I just want a strong intuition as to \*why\* the models work without delving too deep into abstraction. I like theoretical math, but a) I have no talent for math (only improved by doing ungodly amounts of it), b) I will not be getting a PhD. I have been an AI engineer for 2 years now. My background was from software -> AI engineering, and now I have a decent basis for math. Given my goals, is applied math enough, or should I spend more time on theoretical math? Applied Lin Alg 2 vs Regular Regards
document fetched 2026-10-25 00:23 · canonical https://reddit.com/r/deeplearning/comments/1xe20af/theoretic_vs_applied_math_courses_for_gaining