A COllective INtelligence (COIN) is a set of interacting reinforcementlearning (RL) algorithms designed in an automated fashion so that theircollective behavior optimizes a global utility function. We summarize thetheory of COINs, then present experiments using that theory to design COINs tocontrol internet traffic routing. These experiments indicate that COINsoutperform all previously investigated RL-based, shortest path routingalgorithms.
展开▼