摘要：In this article some transport-layer access network deployment scenarios that show user equipment accessing the NGN are considered. The figures used to illustrate these cenarios show physical devices and indicate high-level functionality are shown. In the paper we are stating results of issue for developing: algorithms of analysis of quality and choice the most adequate activate function at neural network for smoothing of non-stationary process; algorithms of control the mistakes in the data of a non-stationary nature at formation of training subset; algorithms of adaptive training of neural networks in view of errors in structural components. The new technique and algorithm for analysis of information dynamic characteristics under their incremental characteristics, and also the general principles and rules for the adaptive control of data by non-stationary nature are developed. The general and private decisions of tasks, formulas for estimation of control optimum borders and of minimal mean square deviation are received for the control of the information at stationary, partial-stationary and non-stationary processes. Quality of Service is the subject of multiple researches in network technologies industry that’s why a group of scientists are always working out the new methods and algorithms, which provide corresponding quality of service. We will study this question from the position of routing protocols in on IP-based networks. An increasing number of emerging Internet applications require better than best effort quality of service (QoS) that is offered by the current Internet. Applications such as voice over IP, video on demand (VoD) need end-to-end QoS guarantees defined in terms of throughput, delay and loss rate. However, QoS guarantee for these services require to match optimal routing protocol. That is can significantly increase the complexity and affect the scalability of the network. This paper addresses the analysis of the backbone routing protocols. This paper focuses on relatively simple direct search methods for solving the optimization problem of minimizing a loss function L = Z (θ ) subject to the parameter vector 0 lying in some set 0.
