After 10 Years Teaching ML Math, He Says You Only Need 20% of University Math

TivadarDanka · x · 2026-09-05

Tivadar Danka, who has spent a decade teaching math to machine learning engineers, published a roadmap arguing that 80% of university math is irrelevant to real ML work. What matters is the core of three pillars: linear algebra (describing models), calculus (fitting models to data, e.g. understanding SGD), and probability theory (prediction under uncertainty). He recommends using the article as a reference map — go deep on each concept, then return to the roadmap — taking beginners from zero to understanding how neural networks work. Published on his Substack, The Palindrome.

Original post →

More from Research

Research channel →