A compilation based approach to inference in dynamic evidential networks

A wide variety of compilation techniques have been proposed in the literature for inference in static graphical models. One of the most widely used approaches is the arithmetic circuit method. Compiling a graphical model into an arithmetic circuit provides a compact representation of the polynomial that it induces. This representation allows linear time exact propagation in the size of the circuit. In this paper, we consider the problem of performing exact inference on dynamic evidential models. We extend the standard arithmetic circuit based approach to inference in dynamic evidential networks.