First Steps

After we have started MTK, our first step will be to check which plugins are available, and if we have one that allows us to work with a hidden Markov model. This can be done with the list command:

Example:

MTK:1> list plugins
-----------------
Available Plugins
-----------------
            example - Example plugin
         floatvalue - Float Value Samples
               ghmm - Hierarchical Gilbert Hidden Markov Model
                hmm - Hidden Markov Model
          hmm_batch - Hierarchical General Hidden Markov Model - Fixed Batch
 hmm_batch_variable - Hierarchical General Hidden Markov Model - Variable Batch
           intvalue - Integer Value Samples

To obtain information about any plugin, or even about some plugin method, use the help command:

help $<$plugin_name$>$
or
help $<$plugin_name$>$.$<$method_name$>$

Example:

MTK:2> help hmm
This plugin defines a discrete-time, discrete-space hidden Markov
model. Both observations and hidden states are non-negative 32-bit
integers.
    Constructors: hmm( )
                  hmm( <N>, <M> )
    Where:
        <N> - number of hidden states.
        <M> - number of observation symbols.

--------------------
Available attributes
--------------------
         N - number of hidden states
         M - number of observation symbols
     pi[i] - initial probability for the i-th state
   A[i][j] - transition probability from state i to state j
   B[i][j] - probability of symbol j at state i
 result[i] - result array of last executed command

------------------
Available displays
------------------
 all - model parameters
  pi - initial state distribution
   A - state transition matrix
   B - observation probability matrix

-----------------
Available methods
-----------------
                  load                  save             normalize
              training              simulate               viterbi
            likelihood              forecast   import_from_tangram
              set_full            set_coxian               set_qbd
           set_gilbert              fix_full            fix_coxian
               fix_qbd           fix_gilbert           set_epsilon
         symb_sum_dist             symb_tavg            state_prob

So far so good! MTK has a plugin that allows us to work with a hidden Markov model - the hmm plugin. Now let's get to work!