Model generation of aggregate ATM traffic using a neural control with accelerated self-scaling

In this paper it is proved that artificial neural networks (ANN) are adequate tools to obtain superposition models of multiplexed Asynchronous Transfer Mode (ATM) traffic sources. In particular we approximate a superposition of individual ATM sources by a two-state Markov Modulated Poisson Process (MMPP). This approximation is performed by a neural system, which consists in an indirect controller that calculates the MAPP models just knowing a traffic descriptor. The learning of the neural networks is achieved through an improved self-scaling algorithm to avoid the need of establishing a range for the input and output variables. Finally, the MMPP approach is evaluated by numerical examples.