GSoC/GCI Archive
Google Summer of Code 2011 Apache Software Foundation

Parallel Viterbi algorithm for Hidden Markov Model

by Sergey Bartunov for Apache Software Foundation

The Viterbi Algorithm is an evaluating algorithm for Hidden Markov Model. It estimates the most likely sequence of hidden states called "Viterbi path" from given sequence of observed states and is also used in Viterbi training. This project is intended bring parallel evaluating functionality for HMM in Mahout and also investigates how such dynamic algorithms could be implemented in MapReduce paradigm.