New answers tagged

0

I learned from this blog: https://vasishth.github.io/bayescogsci/ by Prof. Dr. Shravan Vasishth that In recent years, Bayesian methods have come to be widely adopted in all areas of science. This is in large part due to the development of sophisticated software for probabilisic programming; a recent example is the astonishing computing capability afforded ...


0

You can solve this problem with 2 way. You can set new observations to your HMM model and run forward probabilities.Once you have you should focus on last column of probabilities because the forward algorithm efficiently sums over all the probabilities of all possible paths to each state for each observation in each sequence. The end result is that the log-...


1

Such a state sequence probability is indeed quite low (in the context of your small example of length 4 where there is only 16 possible sequences...). A longer sequence might improve the infered sequence probability (not in absolute value, but opposed to the other sequences' probabilities) because you would generate a chain which reflects a strong markovian ...


1

This is because the inference (estimation of the hidden states) made by the viterbi function optimizes another criterion than the posterior function. With $\pmb{X}$ the vector of hidden random variables and $\pmb{Y}$ the vector of observed random variables, viterbi gives you the Maximum A Posteriori (MAP) estimate defined by: $$ \hat{\pmb{x}}^{MAP} = \mathrm{...


Top 50 recent answers are included