Least-squares support vector machine-based learning and decision making in cognitive radios

摘要

Cognitive radio (CR) can improve system performance and increase its adaptation ability because of its high intelligence in configuring system parameters dynamically. The key of intelligence in CR is its learning capability. After comparing the conventional optimisation decision-making methods and the learning-based ones in intelligent reconfiguration in CR, this study proposes a general learning-based decision-making model framework. According to the framework, a concrete implementation of learning and decision making on the constructed CR communication scenario based on the leastsquares support vector machine is demonstrated in detail. Some results of two simulation experiments show that the system performance can be remarkably improved as the CR system learns more reliable knowledge from more communication instances experienced, and that the generalisation capability of the model is quite good.