Symbolic Function Network

Abstract

Quality of Service (QoS) of telecommunication networks could be enhanced by applying predictive control methods. Such controllers rely on utilizing good and fast (real-time) predictions of the network traffic and quality parameters. Accuracy and recall speed of the traditional Neural Network models are not satisfactory to support such critical real time applications. The Symbolic Function Network (SFN) is a HONN-like model that was originally motivated by the current needs of developing more enhanced and fast predictors for such applications. In this chapter, the authors use the SFN model to design fast and accurate predictors for the telecommunication networks quality control applications. Three predictors are designed and tested for the network traffic, packet loss, and round trip delay. This chapter aims to open a door for researchers to investigate the applicability of SFN in other prediction tasks and to develop more accurate and faster predictors.