1

The hmmlearn reference page and the code mentions the usage of the "verbose" keyword as follows: "verbose (bool, optional) – When True per-iteration convergence reports are printed to sys.stderr. ". During runs, I can see the a 2-column output for every iteration which looks like:

    19       60924.7863        +156.5809
    20       61054.3471        +129.5608

, where 19 and 20 are the iteration numbers. Can you please tell what are the other two numbers in each row? I assume one of them (the first) is the log-likelihood.

1 Answers1

1

The second number on each row should be the likelihood of the model with respect to the training data. The third number is the evolution of the likelihood with respect to the previous iteration.

For instance, for iteration 20, the difference in likelihood with respect to iteration 19 is 61054.3471 - 60924.7863 = 129.5608

Eskapp
  • 3,419
  • 2
  • 22
  • 39
  • Likelihood = the most likely the data could have been generated by the model. In that sense, the higher the better. Up to a certain point however, as if the model fits too well the training data, it may not generalize well to other data (Read about the concept of: Overfitting). However, overfitting is not so much related to the number of iterations but rather to the number of parameters in the model. – Eskapp Jun 21 '19 at 18:42