Probabilistic Finite Domains.
This thesis presents a set of programming constructs that are
capable of modelling probabilistic concepts and of computing with
such concepts. The main objectives are to
provide: a theoretically sound, practically
achievable and notationally intuitive formalism.
The probabilistic programming constructs are presented in
the form of a system called probabilistic finite domains, which
enhances the Logic Programming paradigm with a novel constraint solver.
In doing so, we are able to take advantage of the
knowledge representation power of probability.
In particular we investigate:
first, the duality of the two interpretations of probability
to the problems researchers face when wishing to create a probabilistic
formalism and second, the use of probability as a unifying model
for computational derivations. Some programming examples
and a simple implementation are also described.