pub trait Trainable<Observation> {
fn train_baum_welch(
&self,
observations: &[Vec<Observation>],
n_iter: Option<usize>,
tol: Option<f64>
) -> (Array1<LogProb>, Array2<LogProb>, Array2<LogProb>, Array1<LogProb>);
fn update_matrices(
&self,
transition_hat: Array2<LogProb>,
observation_hat: Array2<LogProb>,
initial_hat: Array1<LogProb>,
end_hat: Array1<LogProb>
);
}Expand description
A trait for trainning Hidden Markov Models (HMM) with generic Observation type using Baum-Welch algorithm.
Required methods
Iterative procedure to train the model using Baum-Welch algorithm given the training sequences.
As arguments, a set of sequences (observations) and two optional argumets: maximum number of iterations (n_iter) and tolerance (tol).
The baum-welch iterative training procedure will stop either if it reaches the tolerance of the relative log-likelihood augmentation (default 1e-6) or
exceed the maximum number of iterations (default 500).
This feature comes in handy in Bam-Welch algorithm when doing an update of HMM parameters.
After receiving the estimated parameters found after trainning, this method updates the values in the HMM model.