jazzparser.misc.raphsto.train.RaphstoBaumWelchUnigramTrainer

Class with methods to retrain a Raphsto model using the Baum-Welch EM algorithm. Special trainer to train unigram models. That is, it doesn't update the transition distribution.

train(self, emissions, max_iterations=None, convergence_logprob=None, logger=None, processes=1, save=True, save_intermediate=False)

source code

Performs unsupervised training using Baum-Welch EM.

This is an instance method, because it is performed on a model that has already been initialized. You might, for example, create such a model using initialize_chord_types.

This is based on the training procedure in NLTK for HMMs: nltk.tag.hmm.HiddenMarkovModelTrainer.train_unsupervised.

Parameters:

emissions (list of lists of emissions) - training data. Each element is a list of emissions representing a sequence in the training data. Each emission is an emission like those used for jazzparser.misc.raphsto.RaphstoHmm.emission_log_probability, i.e. a list of note observations
max_iterations (int) - maximum number of iterations to allow for EM (default 100). Overrides the corresponding module option
convergence_logprob (float) - maximum change in log probability to consider convergence to have been reached (default 1e-3). Overrides the corresponding module option
logger (logging.Logger) - a logger to send progress logging to
processes (int) - number processes to spawn. A pool of this many processes will be used to compute distribution updates for sequences in parallel during each iteration.
save (bool) - save the model at the end of training
save_intermediate (bool) - save the model after each iteration. Implies save

Overrides: RaphstoBaumWelchTrainer.train

Class RaphstoBaumWelchUnigramTrainer

train(self, emissions, max_iterations=None, convergence_logprob=None, logger=None, processes=1, save=True, save_intermediate=False)

update_model(self, model, save=True)

MODEL_TYPES