Analysis, implementation and application of (rule-based) programming languages for probabilistic logic learning.

Machine learning is the subfield of artificial intelligence and computer science that studies how machines can learn from experience to improve their performance on specific tasks. Probabilistic logic learning (PLL) is a recent research area in which probability theory and logic programming are combined for machine learning purposes. Logic programming based approaches have the major advantage of allowing to represent and reason about (uncertain) knowledge in a symbolic and declarative way. The programmer can give a high-level specification of the problem without having to worry about the operational details.This project contributes to PLL research by studying and implementing a new programming language for PLL. This new language, called CHRiSM, unifies several existing PLL approaches in an elegant and natural way. Based on chance rules, which can be interpreted as generalized probabilistic causal processes, CHRiSM has the potential to become a high level framework for combining and extending the existing PLL approaches, as well as an efficient rapid prototyping language for new approaches. Application domains include computational linguistics, bio-informatics, scheduling, spatio-temporal reasoning, robotics, multi-agent systems, the semantic web, type systems, testing and verification, and computational musicology.

Keywords:CHRiSM, Constraint handling rules, Probabilistic logic learning, Machine learning, Declarative languages, Logic programming, Rule-based programming

Disciplines:Applied mathematics in specific fields