Information Cascades in the Flocking Algorithm
Link: http://harry.me/blog/2011/02/17/neat-algorithms-flocking/
An information cascade is often described as a stable network with varying information. It is also possible to have an information cascade with varying nodes. In the flocking algorithm example, we have an information cascade with mobile nodes.
Some background on the flocking algorithm. The flocking algorithm is an algorithm on how animals move in herds or flocks. We call each node a boid, and it can have information on the location, velocity, and direction of the other boids within a radius of our boid. Therefore, we have a network containing nodes(the boids) and information. The behavior of each boid is determined by the boids around it. The behavior is split into three parts, cohesion, alignment, and separation, each one denoting a type of information the network can use as a cascade.
Each behavior is a cascade. For each behavior, the boid will take the information of the neighboring boids and change its movement. Cohesion tells the central boid where to move. The boid calculates the center point of all the boids within a radius and tries to move towards that center direction. Alignment determines where each boid is facing. It calculates the average direction of the neighboring boids, and turns towards that direction. Finally, separation is a mechanism in which boids which are too close to each other separate. These three behaviors make up the information cascade of the flocking algorithm.
This flocking algorithm is a real-life example of the information cascade introduced in lecture. Information about the surrounding boids are transmitted to the center boid, affecting his decision. Like birds in a real flock, the individual decision will have a negligible impact on the whole network of boids. The behavior of individuals in this network is ultimately decided by the neighboring boids any boid currently inhabits.