readme

来自「HMMBOX, version 3.2, William Penny, Impe」· 代码 · 共 45 行

HMMBOX, version 3.2, William Penny, Imperial College, Feb 1999
Matlab toolbox for Hidden Markov Models

(Adapted from Machine Learning Toolbox
Version 1.0  01-Apr-96
Copyright (c) by Zoubin Ghahramani, University of Toronto)

The software uses some NETLAB routines 
(see http://neural-server.aston.ac.uk/netlab/index.html)
so you'll need to have NETLAB on your search path

See the file VERSION for what's new in this version.

The following observation models have so far been implemented:
Gaussian, Gaussian with common covariances, Likelihood, AR.

DEMONSTRATIONS:

demgausshmm.m       uses Gaussian observation model on AR features
demlike.m        the time series values are themselves likelihoods
demar.m          uses AR observation model on original time series

ROUTINES:

hmminit          initialise Gaussian observation HMM
hmmtrain         train HMM
hmmdecode        make classifications using HMM

obsinit          initialise Gaussian observation model
obslike          calculate likelihood of data given observation model
obsupdate        update parameters of observation model

rsum             row sum of matrix
rprod            row product of matrix and vector
rdiv             row division of matrix by vector

To extend the HMM to different observation models add the required
code into obsinit, obslike and obsupdate.

DATA:

Read the file DATA-STRUCTURE.