The forward-backward algorithm is an algorithm for computing posterior marginals in a hidden Markov model (HMM). It is based on dynamic programming, and has linear complexity in the length of the sequence. It is used as a component of several other algorithms, such as the Baum_Welch algorithm and block Gibbs sampling in factorial HMMs.
This concept has the prerequisites:
- multivariate distributions (The forward-backward algorithm is an algorithm for marginalization.)
- hidden Markov models (Forward-backward is an algorithm for inference in HMMs.)
- conditional independence (The justification of the algorithm uses the conditional independence properties.)
Nominated!
Delip Rao: “It’s the meat and potataoes of all the modern machine learning algortihmes / models. And it’s how you train most modern machine learning models.”
