Model Checking the Uncertainties in Software Systems Introduced by Intelligent Components

An intelligent software system consists of intelligent components, fundamentally modeled by neural networks, and traditional software components, in which the intelligent components are responsible to make decisions. Thus, intelligent systems can adapt to environment changes in the running time. However, due to the fact that a neural network highly relies on the training data, its structure and quality are both uncertain. These uncertainties will be passed to the rest of the software system by the intelligent components. This paper presents a framework to model checking such kind of intelligent system software.