Charles Explorer logo
🇬🇧

Machine-learning abstractions for component-based self-optimizing systems

Publication at Faculty of Mathematics and Physics |
2023

Abstract

This paper features an approach that combines machine-learning abstractions with a component model. We target modern self-optimizing systems and therefore integrate the machine-learning abstractions into our ensemble-based component model DEECo.

We further endow the DEECo component model with abstractions for specifying self-optimization heuristics, which address coordination among multiple components. We demonstrate these abstractions in the context of an Industry 4.0 use case.

We argue that incorporating machine learning and optimization heuristics is the key feature for modern smart systems, which learn over time and optimize their behavior at runtime to deal with uncertainty in their environment.