Robust Training of a link adaptation cognitive engine

In this paper, we provide a new perspective and insight into the process of finding the maximal performing method using a Cognitive Engine (CE) for link adaptation. It is found that near maximal performance can be reached relatively fast, even when a small number of the available communication methods provide adequate performance. The parameters that affect the expected number of trials are fully discussed along with analytical and simulation results. Finally, we provide the novel Robust Training Algorithm (RoTA), which given at least one method that exceeds the minimum performing requirements, adaptively maintains a communication link with the minimum required performance. The RoTA allows the CE to both continue learning and maintain a stable link for mission-critical applications.