mlpy.stats.dbn.hmm.HMM¶
-
class
mlpy.stats.dbn.hmm.
HMM
(ncomponents=1, startprob_prior=None, startprob=None, transmat_prior=None, transmat=None, emission_prior=None, emission=None, n_iter=None, thresh=None, verbose=None)[source]¶ Bases:
mlpy.optimize.algorithms.EM
Hidden Markov Model base class.
Representation of a hidden Markov model probability distribution. This class allows for easy evaluation of, sampling from, and maximum-likelihood estimation of the parameters of a HMM.
See the instance documentation for details specific to a particular object.
Parameters: ncomponents : int
Number of states in the model.
startprob_prior : array, shape (ncomponents,)
Initial state occupation prior distribution.
startprob : array, shape (ncomponents,)
Initial state occupation distribution.
transmat_prior : array, shape (ncomponents, ncomponents)
Matrix of prior transition probabilities between states.
transmat : array, shape (ncomponents, ncomponents)
Matrix of transition probabilities between states.
emission : cond_rv_frozen
The conditional probability distribution used for the emission.
emission_prior : normal_invwishart
Initial emission parameters, a normal-inverse Wishart distribution.
n_iter : int
Number of iterations to perform during training, optional.
thresh : float
Convergence threshold, optional.
verbose : bool
Controls if debug information is printed to the console, optional.
Examples
>>> from mlpy.stats.dbn.hmm import GaussianHMM
>>> model = GaussianHMM(ncomponents=2, startprob_prior=[3, 2])
Create a gaussian hidden Markov model
>>> import scipy.io >>> mat = scipy.io.loadmat('data/speechDataDigits4And5.mat')) >>> x = np.hstack([mat['train4'][0], mat['train5'][0]])
Load data used for fitting the HMM and fit the HMM:
>>> model.fit(x, n_init=3)
Attributes
startprob_prior
Vector of initial probabilities for each state. transmat_prior
Transition probability matrix. ncomponents (int) The number of hidden states. nfeatures (int) Dimensionality of the Gaussian emission. startprob (array, shape (ncomponents,)) Initial state occupation distribution. transmat (array, shape (ncomponents, ncomponents)) Matrix of transition probabilities between states. emission_prior (normal_invwishart) Initial emission parameters, a normal-inverse Wishart distribution. emission (cond_rv_frozen) The conditional probability distribution used for the emission. Methods
decode
(obs[, algorithm])Find the most likely state sequence. fit
(obs[, n_init])Estimate model parameters. predict_proba
(obs)Compute the posterior probability for each state in the model. sample
(length[, size])Generates random samples from the model. score
(obs)Compute log probability of the evidence (likelihood) under the model. score_samples
(obs)Compute the log probability of the evidence.