Neural Network Topological Evolvement
Yuan Zhongda, Zhen Ye · 2006
In order to obtain instance response from evolving neural networks, we separate the evolving procedure and the output procedure in the traditional GA approach, and make them simultaneously. Thanks to the support from database system, this modified GA has three characteristics, namely serialized storage, persistent evolving and instance response to end user's request. In our experiment, this approach shows off better performance over the conventional strategy