GHMM training

[renatocf]

Implement segseq logic inside GHMM static method train and creator.

The intended interface is:

/* Client code */
auto ghmm_trainer = GHMM::labelingTrainer(state_trainers);
ghmm_trainer->addTrainingSet(seq1);
ghmm_trainer->addTrainingSet(seq2);
...
ghmm_trainer->addTrainingSet(seqN);
ghmm_trainer->train(algorithm_tag, ...);

/* GHMM::train code */
{
  for (auto labeled_seq : split_sequence(ghmm_trainer->training_set())) {
    ghmm_trainer->state(labeled_seq.index())->add_training_set(labeled_seq.sequence());
  }
  for (auto state_trainer : ghmm_trainer->trainers()) {
    state_trainer->train(...) /* How to pass these parameters? */
  }
  /* train ghmm transitions */
}

[renatocf]

@igorbonadio,

So, I figured out an idea that can solve our problem. It will allow us to keep the method train receiving parameters, without (explicitly) creating any parameter objects. However, I need to know: how will trainers (#10) communicate with ToPS language (#11) to create a GHMM? How a tops::lang AST will get values at runtime and pass them to the different train methods?

Before making further implementations, I believe we should think a little bit more about how both things will dialogue.