Neural Network Approximation of Characteristics of Multi-Channel Non-Markovian Queuing Systems

It is proposed to use a neural network to calculate an approximation of the probabilistic-time characteristics of multichannel queuing systems (QS) with a "warm-up" and the unlimited capacity of the queue. From the results of numerical experiments, we observe a significant reduction in the complexity of computing probabilistic-time characteristics of the multi-channel QS with "warm-up" with minor errors of calculation of characteristics, compared with the numerical iterative algorithms. The advisability of the use of Bayesian regularization method for training a neural network and the best number of neurons are shown.