Continuous-Time Bayesian Networks

A CTBN with its CIMs

A CTBN(Continuous-Time Bayesian Networks) is a graphical model that allows a Bayesian Network to evolve over continuous time. This pyAgrum library offers ways to create such models, to have a graphical representation but also to learn such models (the dependency between variables and their distribution parameters) using exact inference and sampling as well. To this day Forward Sampling is the only sampling method available.

The goal is to have the properties of a discrete Markov Chain but at continuous time, which means that a random variable is allowed to switch state at any time. To depict the time a variable spends in a state before switching to another, we use an exponential distribution.

Tutorials

Reference