论文部分内容阅读
本文在研究神经网络自组织方法的基础上,通过模拟生物群落演化的动态机制,提出了一种新的自组织模型GGM,在GGM中,我们利用聚类中心的动态增生和消亡,使自组织学习调整直接局部化到单个聚类中心,从而避免了全局一局部调整的模拟退火训练过程,大大地提高了自组织学习的速度和适应能力.
On the basis of studying the self-organizing neural network method, a new self-organizing model GGM is proposed by simulating the dynamic mechanism of the evolution of biological communities. In GGM, we use the dynamic growth and disappearance of clustering centers, Learning adjustment directly localized to a single cluster center, thus avoiding the global one part of the adjustment of simulated annealing training process, greatly improving the speed of self-organized learning and adaptability.