Hi , I wonder to know Deep Markov model tutorial use for loop to implement , but HMM tutorial use pyro.markov . Both of implement are same markov independent ? Thanks !
You’re correct that the models have the same conditional independence structure. However, the information about a model conveyed by
pyro.markov is only used by inference algorithms that use Pyro’s enumeration machinery, namely variational inference with
TraceEnum_ELBO and MCMC with
HMC. Since the DMM example does not use one of those algorithms, using
pyro.markov in it is unnecessary.
Thanks a lot !!!