Neutral Networks In Artificial Evolution

Neutral network basically mean a set of genotypes with the same fitness value. Consider the figure above which I've plagiarized from Inman's article in scholarpedia. The yellow, green and blue layers are the neutral networks and there is a single fitness value for each layer. The red bars represent the mutation which allows movement from one network to another.

So what is the significance of this??

The most important outcome of the existence of the above phenomenon is in the effect it has on the design of the genetic algorithm. Classical view of genetic algorithm has always neglected this phenomenon as a result of which genetic algorithm look the way they do today. In the presence of the neutrality, the design of genetic algorithm changes as pointed out by Lionel Barnett in the paper- Lionel Barnett (2000) Netcrawling - Optimal Evolutionary Search with Neutral Networks. This paper presents a new algorithm to deal with problems containing neutrality.

A good description of neutral network by Inman Harvey is available here. At the time of this post, this is the only proper description of neutral network available on the net.