Baum-Welch Algorithm

The Baum-Welch algorithm is a special case of the expectation-maximization algorithm used to find the unknown parameters of a hidden Markov model, applied across electrical engineering, statistical computing and bioinformatics. It uses the forward-backward algorithm to compute the statistics needed for the expectation step and is the primary method used for training hidden Markov models, though it is numerically unstable in its basic form because the joint probabilities it computes recursively can shrink below machine precision as the number of variables grows.