Expansion of neural networks and the learned techniques

We study the development of the capability of expansion of an intelligent system structure as well as the learned techniques. The learning structure consists of a pool of neurons. For a primary skill, a subset of neurons is used. A genetic algorithm (GA) is adopted to search for adequate connections and weights for performing the skill. To develop more advanced techniques, unused neurons are selected. The GA is then used to search for connections within the new set of neurons and between the newly selected neurons and the old ones. The original structure is not altered. This preserves the previously learned skills and the new skill can be established based on existent ones. It is expected that, using such hierarchical learning, learning and skill growing can be more efficient. Although the purpose of this study is to resolve the problem of skill expansion upon completing design, it is noted that the scheme can also be applied to relax learning difficulty by decomposing a difficult skill