Ticket #3726: sage-3726-part7.patch

File sage-3726-part7.patch, 1.8 kB (added by was, 5 months ago)

  • a/sage/stats/hmm/chmm.pyx

    old new  
    596596            exception on error 
    597597 
    598598        EXAMPLES: 
     599        We train a very simple model: 
     600            sage: m = hmm.GaussianHiddenMarkovModel([[1]], [(0,1)], [1]) 
     601            sage: m.baum_welch([1,1,1,1]) 
    599602 
    600         NOTE: Training for models including silent states is not yet supported. 
    601  
    602         REFERENCES: 
    603             Rabiner, L.R.: "`A Tutorial on Hidden Markov Models and Selected 
    604             Applications in Speech Recognition"', Proceedings of the IEEE, 
    605             77, no 2, 1989, pp 257--285.  
     603        Notice that after training the mean in the emission parameters changes 
     604        form 0 to 1. 
     605            sage: m 
     606            Gaussian Hidden Markov Model with 1 States 
     607            Transition matrix: 
     608            [1.0] 
     609            Emission parameters: 
     610            [(1.0, 0.0001)] 
     611            Initial probabilities: [1.0] 
    606612        """ 
    607613        cdef ghmm_cmodel_baum_welch_context cs 
    608614 
    609615        cs.smo      = self.m 
    610616        cs.sqd      = to_cseq(training_seqs) 
     617        cs.logp     = <double*> safe_malloc(sizeof(double)) 
    611618        cs.eps      = log_likelihood_cutoff 
    612619        cs.max_iter = max_iter 
    613620         
    614621        if ghmm_cmodel_baum_welch(&cs): 
    615622            raise RuntimeError, "error running Baum-Welch algorithm" 
    616         #ghmm_cseq_free(&cs.sqd) 
     623        ghmm_cseq_free(&cs.sqd) 
     624         
    617625         
    618626cdef ghmm_cseq* to_cseq(seq) except NULL: 
    619627    """