YAPOHMM

[Andreas Stolcke]

(Yet another paper on Hidden Markov Models)

The following paper, to appear in NIPS-5, is now available by FTP from
ftp.icsi.berkeley.edu (128.32.201.7) in the file /pub/ai/stolcke-nips5.ps.Z.

Since this is only the latest in a series of similar announcements on 
connectionists I will spare you the ftp instructions.  Let me know if
you don't have ftp access and want an e-mailed copy.

-----

Hidden Markov Model Induction by Bayesian Model Merging

Andreas Stolcke and Stephen Omohundro

This paper describes a technique for learning both the number of
states and the topology of Hidden Markov Models from examples. The
induction process starts with the most specific model consistent with
the training data and generalizes by successively merging states.
Both the choice of states to merge and the stopping criterion are
guided by the Bayesian posterior probability.
We compare our algorithm with the Baum-Welch method of estimating fixed-size
models, and find that it can induce minimal HMMs
from data in cases where fixed estimation does not converge or
requires redundant parameters to converge.


--Andreas