Sustainable and Trustworthy Edge Machine Learning

Nowadays, our world is driven by complex, large scale, yet tactile information systems requiring various degrees of trustworthiness. Trustworthiness of the systems always comes with costs. The traditional and rather costly way to understand the behavior of large scale systems is to develop powerful mathematical abstractions that allow us to condense these behaviors and to reason about them at a very abstract level. In our FWF funded project Rucon, we introduce an orthogonal, data driven, and probabilistic concept to model and reason uncertainty of the systems. In Rucon, deliberated system failures are tolerated due to the benefits of the costs and sustainability. Rucon's approach targets large scale near real-time systems like live video analytics, streaming, vehicular applications, and smart city information systems.