Real-Time Tracking of Packet-Pair Dispersion Nodes Using the Kernel-Density and Gaussian-Mixture Models
Mar. 25, 2009 to Mar. 27, 2009
DOI Bookmark: http://doi.ieeecomputersociety.org/10.1109/UKSIM.2009.74
A brief simulation study of real-time packetdispersionmode-tracking using the Gaussian-MixModel (originally devised for real-time backgroundclassification in moving pictures) and an adaptationof the Kernel-Density Estimator is presented. Thesimulated environment consisted of two FIFO storeand-forward nodes where the probe packets interactwith Poisson and Pareto-generated cross-trafficwith a range of packet sizes. The two modelsproduced broadly similar results, able to track nodeactivity under the dynamically changing conditionsassociated with the Pareto cross-traffic. TheGaussian model sometimes replaced the primarymode with a double peak, which disappeared whensome of the model’s parameters were changed.
Bandwidth Measurement, Machine Learning
M. Hosseinpour, M.J. Tunnicliffe, "Real-Time Tracking of Packet-Pair Dispersion Nodes Using the Kernel-Density and Gaussian-Mixture Models", UKSIM, 2009, Computer Modeling and Simulation, International Conference on, Computer Modeling and Simulation, International Conference on 2009, pp. 548-552, doi:10.1109/UKSIM.2009.74