Logprob: probabilistic logic --- foundations and computational applications

Abstract

Among the several possible forms of rational reasoning, logical reasoning and probabilistic inference have received particular attention in practive, having been incorporated both in modeling situations of the world and in computer programs aimed at performing a variety of tasks. However, the possible interactions between these forms of reasoning are not very well understood yet, both in theory and in the practice of constructing computational applications. In the theoretical perspective, it is still to be established the relationship between truth values and probability distributions over formulas. In the practical perspective, it is still to be established the limits of when to apply logical reasoning and when to apply statistic inference. This research project intends to investigate those questions, both in the theoretical and practical perspectives. The project will focus on four interface areas between logic and probabilities, namely, probabilistic satisfiability, probabilistic description logics applied to the semantic web, probabilistic planning and scene analysis. This project intends to produce several pieces of software covering the area under investigation that will be made available under a FLOSS (Free/Libre/Open Source Software) license as part of the activities of the Artificial Intelligence division of the Center for Competence in Free Software at the University of Sao Paulo. (AU)