A Unified View to Machine Learning and Control for Measurement-based Equivalent Bandwidth

This paper outlines a unified view of machine learning and control for the optimization of a communication system. The problem of equivalent bandwidth is taken as a reference. A dedicated classification technique is used to derive insights into the structure of the problem by means of boolean rules over the variables of the system. The approach is of particular interest for many settings in which only measurements of the performance are available. Simulations corroborate the quality of the proposed technique.