Open nutterb opened 9 years ago
Here is an excellent paper that is worth understanding. They implement a Bayesian non-linear state space model using a small dataset and WinBUGS. I think this could be represented as a temporal Bayesian network, composed of two subnetwork (HydeNet) model objects - the one mapping time 0 to time 1, and the one mapping time t to time t+1.
http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.365.9015&rep=rep1&type=pdf
In case the link doesn't work, the article is:
Meyer, Renate, and Russell B. Millar. "BUGS in Bayesian stock assessments." Canadian Journal of Fisheries and Aquatic Sciences 56.6 (1999): 1078-1087.
After reflecting on that paper, I'm getting a better idea of how temporal models with HydeNet should look. Instead of writing about it here, I'll start a vignette.