Self-organization and Intelligence

Qiangfu Zhao

Self-organization (SO) is one of the key issues for generating true intelligence. Without SO, human brains and human societies will just be chaotic systems. The basic SO mechanism is the Leader-Following (LF) behavior. With LF behavior, a complex system can become very well organized after many generations. Without LF behavior, however, even a relatively small dynamic system can become intractable. In fact, the basic learning rule in the well-known Self-Organizing Map (SOM) algorithm is LF. Here, for any given data point, neurons close to the winner tries to imitate the winner, so that they can behave in a similar way for similar data. The basic learning rule in Particle Swarm Optimization (PSO) is also LF. In PSO, each particle tries to imitate the behavior of the local or global leader, while trying to preserve its own search history. Thus, although SOM and PSO are different, both of them are LF based SO algorithms.