CMSC 828A: Fantastic Machine Learning Paradigms and Where to Use Them
Graduate course touring machine learning paradigms — supervised to reinforcement learning, single-task to meta-learning, single-model to multi-agent learning — and where to use each.
Instructor: Tianyi Zhou
Term: Spring
Location: University of Maryland, College Park
CMSC 828A provides an in-depth tour of machine learning paradigms — their assumptions, formulations, motivations, where to use them, and the ideas connecting them. The course follows three routes: (1) supervised → semi-supervised → unsupervised and self-supervised → online and reinforcement learning; (2) single-task → transfer → multi-task → meta-learning; (3) single-model → adversarial → ensemble → federated → multi-agent learning. I developed the whole course, project topics, and reading list from scratch.