Cognitive Networks: Towards Self-Aware Networks

5.1. Introduction

Clark, Partridge, Ramming, and Wroclawski ^{[] [] []} recently proposed a new vision for computer network management – the Knowledge Plane – that would augment the current paradigm of low-level data collection and decision making with higher level processes. One key idea is that the Knowledge Plane would learn about its own behavior over time, making it better able to analyze problems, tune its operation, and generally increase its reliability and robustness. This suggests the incorporation of concepts and methods from machine learning ^{[] []}, an established field that is concerned with such issues.

Machine learning aims to understand computational mechanisms by which experience can lead to improved performance. In everyday language, we say that a person has 'learned' something from an experience when he can do something he could not, or could not do as well, before that experience. The field of machine learning attempts to characterize how such changes can occur by designing, implementing, running and analyzing algorithms that can be run on computers. The discipline draws on ideas from many other fields, including statistics, cognitive psychology, information theory, logic, complexity theory and operations research, but always with the goal of understanding the computational character of learning.

There is general agreement that representational issues are central to learning. In fact, the field is often divided into paradigms that are organized around representational ...