Abstract
A swarm may be defined as a population of interacting elements that is able to optimize some global objective through collaborative search of a space (Kennedy, 2001). The elements may be very simple machines or very complex living beings, but there are two restrictions to be observed: They are limited to local interactions; usually the interaction is not performed directly but indirectly through the environment. The property that makes swarms interesting is their self-organizing behaviour; in other words, it is the fact that a lot of simple processes can lead to complex results.
Key Terms in this Chapter
ACO (Ant Colony Optimization): A swarm intelligence- based optimization heuristic in which small ant-like agents interact indirectly only through changes to the environment and reacting to those changes.
Distance Vector Routing: A routing algorithm that requires that each router simply inform its neighbours of its routing table. The opposite is a link-state routing algorithm.
Collaborative Sequencing: An application in which the sequences of activities followed by some users is used to infer the best sequence for other users.
Stagnation: Situation in which a system does not evolve towards a better solution because it has found a local optimum. For example, sometimes ant colonies stick to existing trails to food even if better, new ones appear because the pheromone trail is so strong they cannot leave it.
Stigmergy: Indirect communication process in which the individual parts of a system communicate with one another, not directly but by modifying their local environment.
Routing: The distributed activity of building and using tables at the nodes of a packet network that state the path to be followed by packets going from a source to a not-directly-connected destination.
Link-State Routing Algorithm: A routing protocol that requires each router to maintain at least a partial map of the network. The opposite is a distance-vector routing algorithm.
Collaborative Filtering: An application in which different users express their individual preferences about some items, and the emerging result is the possibility of making predictions about items not rated or for new users.