Network Congestion: An Example of Non-Linear Extrapolation

Abstract

As more and more people use computers to communicate, network congestion becomes a more and more serious problem. Ideally, we should be able to prevent congestion by predicting its possibility and undertaking the appropriate actions. For predicting, we need extrapolation. It turns out that for network congestion, the linear extrapolation tools that we have considered so far (i.e., extrapolation tools based on families of functions {C₁ · f₁ (x) +...+ C _m · f _m (x)} that linearly depend on the parameters C_i do not work very well. In this lesson, we show that continuous mathematics helps to select non-linear extrapolation tools that lead to a better congestion control. Mathematical methods used in the design of these non-linear extrapolation tools will be used in the following lessons to describe neural networks and fuzzy control.