Statistical Relational Learning for Predictive Maintenance.

Within the field of Artificial Intelligence, there is a lot of interest in combining
probability and expressive representations for dealing with complex relational
and dynamic domains. A common approach to reason with these representations
is to rely on existing techniques for propositional models and thus requires to
first ground the underlying model into a propositional representation. This
strategy comes at a cost, however, as capturing the semantics of the original
representation might lead to a combinatorial explosion and quickly renders
inference intractable. This dissertation investigates how weighted model counting
and knowledge compilation can be used to directly perform inference on the
original representation.
This thesis has three main contributions. First, we propose an exact probabilistic
inference algorithm for propositional dynamic domains. Our approach allowsto exploit different types of structures by compiling the transition model into
an efficient circuit representation. Second, we propose an anytime probabilistic
inference algorithm for relational domains. An efficient circuit representation is
compiled in an incremental way and, at any time in the process, hard bounds
on the inferred probabilities can be computed. Third, we deal with relational
dynamic domains by combining principles from the first and second contribution.
In addition, our approach exploits the given observations to further scale-up
inference.
The techniques presented in this dissertation are evaluated empirically on
various real-world domains and applications such as biological and social network
analysis, web-page classification, electronic circuit diagnosis and game playing.
They outperform state-of-the-art approaches on these problems with respect to
time, space and quality of results.