poisson_hmm_nloglikeobs.py

    def nloglikeobs(self, params):
        #Reconstitute the q and beta matrices from the current values of all the params
        self.reconstitute_parameter_matrices(params)

        #Build the regime wise matrix of Poisson means
        self.compute_regime_specific_poisson_means()

        #Build the matrix of Markov transition probabilities by standardizing all the q values to
        # the 0 to 1 range
        self.compute_markov_transition_probabilities()

        #Build the (len(y) x k) matrix delta of Markov state probabilities distribution. k state
        # probabilities corresponding to k regimes, times, number of time steps (i.e. observations)
        self.compute_markov_state_probabilities()

        #Let's increment the iteration count
        self.iter_num=self.iter_num+1

        # Compute all the log-likelihood values for the Poisson Markov model
        ll = self.compute_loglikelihood()

        #Print out the iteration summary
        print('ITER='+str(self.iter_num) + ' ll='+str(((-ll).sum(0))))

        #Return the negated array of  log-likelihood values
        return -ll