SAM: Support Vector Machine Based Active Queue Management

Muhammad Saleh Shah,Asim Imdad Wagan,Mukhtiar Ali Unar
DOI: https://doi.org/10.48550/arXiv.1604.00557
2016-04-03
Abstract:Recent years have seen an increasing interest in the design of AQM (Active Queue Management) controllers. The purpose of these controllers is to manage the network congestion under varying loads, link delays and bandwidth. In this paper, a new AQM controller is proposed which is trained by using the SVM (Support Vector Machine) with the RBF (Radial Basis Function) kernal. The proposed controller is called the support vector based AQM (SAM) controller. The performance of the proposed controller has been compared with three conventional AQM controllers, namely the Random Early Detection, Blue and Proportional Plus Integral Controller. The preliminary simulation studies show that the performance of the proposed controller is comparable to the conventional controllers. However, the proposed controller is more efficient in controlling the queue size than the conventional controllers.
Networking and Internet Architecture,Machine Learning
What problem does this paper attempt to address?