Expectation Maximization (EM): a Mini-Tutorial
By Bob Woodham
Expectation Maximization (EM) is used in many areas, including computational
vision, to estimate parameters of probabilistic models.
This mini-tutorial examines:
1. How Expectation Maximization works (in theory) 2. How Expectation
Maximization is used (in practice) 3. When is Expectation Maximization (or an
approximation thereof)
the method of choice, compared to alternatives
The objective of the mini-tutorial is to broaden our collective understanding of
algorithms based on EM.