- Probabilistic models, such as hidden Markov models or Bayesian networks, are commonly used to model biological data.
- But models are models, which means we need to first choose the model type, and then estimate the parameters based on the observations.
- The expectation maximization algorithm is a natural generalization of maximum likelihood estimation to the incomplete data case.
- other numerical optimization techniques, such as gradient descent or Newton-Raphson, could be used instead of expectation maximization; in practice, however, expectation maximization has the advantage of being simple, robust and easy to implement.
The expectation maximization algorithm enables parameter estimation in probabilistic models with incomplete data.
Reference:
http://ai.stanford.edu/~chuongdo/papers/em_tutorial.pdf
https://math.stackexchange.com/questions/25111/how-does-expectation-maximization-work
https://stats.stackexchange.com/questions/235070/relation-between-map-em-and-max-likelihood
https://www.cs.utah.edu/~piyush/teaching/EM_algorithm.pdf
https://people.eecs.berkeley.edu/~pabbeel/cs287-fa13/slides/Likelihood_EM_HMM_Kalman.pdf