Publication (Journal publications)

Uncertainty Quantification in Mathematics-Embedded Ontologies Using Stochastic Reduced Order Model

, ; , ; Eckert, Martina; ,

Abstract:

To resolve one of uncertainty features, randomness, in ontologies, this paper shows how to characterize uncertainty of concepts from a statistical viewpoint. In addition, with a focus on indirect entities, which are computed from direct entities through mathematical models, the uncertainties propagated from those direct entities are important and should be quantified. Thus, a novel algorithm, named Stochastic Reduced Order Model (SROM), is presented to be applied to quantify the ontological uncertainty propagation in presence of multiple input entities. This SROM-based method could approximate the statistics of indirect entities accurately and efficiently by using a very small amount of samples of input entities. The computational cost is considerably reduced while guaranteeing a reasonable degree of accuracy. Furthermore, the predicted statistics of output entities could be regarded as high-level information and be beneficial for other ontological operations, such as ontology filtering and ontology reasoning. The implementation of the SROM algorithm is non-intrusive to the mathematical model; therefore, this algorithm could be applicable to quantify uncertainty in ontologies with any mathematical relationships.

Research areas:

• Natural interfaces for therapeutic ends

Year:

2017

Type of Publication:

Journal publications

Journal:

IEEE TRANSACTIONS ON KNOWLEDGE AND DATA ENGINEERING

Volume:

29

Number:

4

Pages:

912-920

DOI:

10.1109/TKDE.2017.2651819