Author:

Keywords:

probabilistic logic programming, datasets, probabilistic graphs, detail abstraction, applications, probabilistic dictionary, c. elegans neural network, social network

Abstract:

Deriving knowledge from real-world systems is a complex task, targeted by many scientific fields. Such systems can be viewed as collections of highly detailed data elements and interactions between them. The more details the data include the more accurate the system representation is but the higher the computational requirements become. Using abstractions to summarize details is a well-known technique. Abstraction often leads to an accurate model of a system but, in other cases, introduces inaccuracies that we want to quantify. In this paper we propose an approach based on different levels of abstraction to define different representational models for a system. We use probabilities to quantify the inaccuracies that are introduced during the abstraction process. Such models then are used for reasoning and learning in a probabilistic environment. We use three example datasets (a small and simple social network, a probabilistic dictionary of approximately 300 words and a real biological neural network) to support our abstraction approach in probabilistic context. The environment we use is ProbLog - a small but powerful probabilistic extension of Prolog.