The hmm_batch plugin has two different constructors:
hmm_batch( )
hmm_batch( N, M, B )
where the first one creates an empty hierarchical general hidden Markov model
(it assumes its parameters will be loaded hereafter); and the second one creates
a HMM-Batch with N hidden states, M observation symbols,
whose observation batch size is B, and initializes its parameters
with random values, satisfying the stochastic constrains. Each of the states
is associated with a unique integer number, ranging from to , which
identifies the state. The observations symbols, emitted by the model, range from
to , and each lower-level state, inside a hidden state , is
associated with one, and only one, symbol.