Math in Machine Learning

This course builds the mathematical foundations essential for understanding classical and modern machine learning methods from the ground up. The focus will be on mathematical formalism and intuition rather than applications per se, however, undergirded ML methods and algorithms will be highlighted and broadly frame the topics covered and their arc: in algorithms that learn from data, we must define a representation space (linear algebra), orient ourselves (calculus), model things (probability), then learn (statistics) about our dataset using our computational tools (numerical methods). We will survey some of the most relevant ideas from these broad areas of math, including linear transformations, eigenvalues and eigenvectors, singular value decomposition, gradients, Jacobians, Hessians, random variables, probability distributions, Bayes’ Theorem, pseudo-random numbers, autodifferentiation, and numerical error. Students completing DATA1010 will be prepared for success in subsequent ML courses like DATA2060, DATA2080/CSCI1470, and DATA2100.