susierao / DiscreteFHMM

Factorial Hidden Markov Models (FHMM) implemented for discrete observation values (like words)

Implementation of Factorial Hidden Markov Model (FHMM) for discrete-valued observation case (eg. words). Features:

• New variational algorithm for learning the parameters of FHMM for discreted-valued observation cases
• Faster training by multi-threading
• Faster training by online learning (training in mini-batches)

Running

Simply run: ./run.sh in the base directory

Modify config.Config file to use your own data