Expectation Maximization - Introduction and Examples

Learn the fundamentals of the Expectation Maximization (EM) algorithm through this 38-minute lecture from the Computational Genomics Summer Institute. Explore the theoretical foundations of EM as a powerful statistical method for finding maximum likelihood estimates when dealing with incomplete data. Discover how this iterative algorithm alternates between expectation and maximization steps to converge on optimal parameter estimates. Examine practical applications and worked examples that demonstrate the algorithm's utility in computational genomics and related fields. Gain insights into the mathematical framework underlying EM, including its relationship to maximum likelihood estimation and its convergence properties. Study key research contributions that shaped the development of EM algorithms, from the foundational 1977 paper by Dempster, Laird, and Rubin to modern extensions and variations. Understand connections to MM (Majorization-Minimization) algorithms and explore real-world applications such as emission tomography reconstruction that showcase the algorithm's versatility across different domains.