Cognitive Networks: Towards Self-Aware Networks

9.5. Design Decisions Impacting Learning and Reasoning

Regardless of the learning and reasoning methods employed by the cognitive network, there are several design decisions that impact the network's performance. These decisions provide a set of 'axes' upon which different designs can be classified and categorized independent of the cognitive process implementation. Understanding how these decisions affect various network objectives can lead to sets of design space trade-offs, providing the network engineer with guidelines for selecting and incorporating specific cognitive features ^[].

Referring to Figure 9.1, cognitive elements are the distributed decision-making and implementation entities in the network and act as part of a larger cognitive process that attempts to attain a set of end-to-end goals that are defined by the operators, users, applications or resources of the network. Cognitive elements in many instances meet the definition for an agent when they are situated, autonomous and flexible. For instance, in the context of a wireless network, cognitive networks may consist of multiple cognitive radios with the cognitive radios acting as the cognitive elements. These radios could be described as agent-like. However, cognitive elements may not always have so direct a hardware mapping. For instance, several network devices may distribute the operation of a single cognitive element, such as the case where a cognitive element utilizes a distributed reasoning algorithm. In these ...