An optimization algorithm often used for maximum-likelihood estimation in the presence of missing data.
The expectation maximization (EM) algorithm is an optimization algorithm. It is an instance of a broader class of optimization algorithms known as Majorization-Minimization (MM). It is a cornerstone of statistics since it particularly suitable for "skipping" local maxima which often arise in likelihood maximization problems, especially in the presence of missing data.
More specifically, the EM algorithm is an iterative method for finding maximum likelihood estimates. The typical form of the EM algorithm works as follows:
- Expectation Step: Compute the expected value of the log-likelihood function based on the current estimate.
- Maximization Step: Update the estimate by maximizing the result of the expectation step.
