I’m building a time series model with auto-regressive components (a variant of GARCH).
From the sparse documentation on the markov handler, it looks like it might be helpful. Are there any examples illustrating its use as an iterator, and accessing the history from within it?
Hi @amos, for example usage you can check out the hmm example. Note that pyro.markov does not store anything, rather it merely allows you to tell Pyro that in your model, the dependency structure of a time series does not let a variable depend directly on another variable longer than history ago. You would need to store such variables in your own Python data structure to do that.
Huh… I’m curious what does it iterate through then?
I get it, it just passes through an iterator…