Abstract: Dynamic Bayesian networks (DBNs) are well suited to the modeling of temporally variable properties of computer users, such as time pressure and cognitive load. One challenge is to develop new methods for limiting the computational complexity of DBNs, so that they can be applied to real-time user modeling. A second goal is to design and test appropriate structures and preprocessing methods for DBNs that interpret data from a user's motor behavior and speech, as well as physiological data.
Abstract: We extend the differential approach to inference in Bayesian networks (BNs) to handle specific problems that arise in the context of dynamic Bayesian networks (DBNs). We first summarize Darwiche's approach for BNs, which involves the representation of a BN in terms of a multivariate polynomial. We then show how procedures for the computation of corresponding polynomials for DBNs can be derived. These procedures permit not only an exact roll-up of old time slices but also a constant-space evaluation of DBNs. The method is applicable to both forward and backward propagation, and it does not presuppose that each time slice of the DBN has the same structure. It is compatible with approximative methods for roll-up and evaluation of DBNs.
Abstract:
We show in this work an example of how physiological
data of a user acquired by bio sensors and interpreted by
dynamic Bayesian networks on a personal mobile device
are used to adapt notifications to the user state. The
described sample application is an instance of our framework
that is currently under development in our research group,
aiming towards a modular and cross-platform toolkit for
sensor integration.
We will give a brief overview of the whole system with
its components and conclude with an outlook on future
work in this project.