1

Dagum developed DBNs to unify and extend traditional linear state-space models such as Kalman filters, linear and normal forecasting models such as ARMA and simple dependency models such as hidden Markov models into a general probabilistic representation and inference mechanism for arbitrary nonlinear and non-normal time-dependent domains.

I've come across time-series models, HMMs, and Kalman filters independently. But the language used seems to be somewhat "siloed" in the sense that it's not obvious how these models are related in a broader sense. From the wiki quote above, Dynamic Bayesian Networks unify these different models into a cohesive philosophy (for lack of a better word.)

This sounds great! But I'd really like a textbook or online course to help me get up to speed. Could anyone link such a resource?

jbuddy_13
  • 1,578
  • 3
  • 22

1 Answers1

1

An excellent, approachable review that unifies these approaches is A Unifying Review of Linear Gaussian Models by Roweis and Ghahramani, published in Neural Computation. The first two sentences of the abstract are

Factor analysis, principal component analysis, mixtures of gaussian clusters, vector quantization, Kalman filter models, and hidden Markov models can all be unified as variations of unsupervised learning under a single basic generative model. This is achieved by collecting together disparate observations and derivations made by many previous authors and introducing a new way of linking discrete and continuous state models using a simple nonlinearity.

While it doesn't go into DBNs, it does cover HMMs, Kalman filters, and Bayesian networks in general.

Arya McCarthy
  • 6,390
  • 1
  • 16
  • 47