摘要
探讨了知识可增殖人工神经网络的可能的实现途径,利用信息几何进行了理论分析并指出只有利用群体网络才可能解决这一问题。在此基础上提出了一种类图书馆知识可增殖神经网络及其组织算法,该群体网络系统,能够有效地继承其中个体网络的知识,具有知识增殖能力。仿真结果表明了其可行性和有效性。
Using information geometry, the probable realized methods of knowledge-increasable artificial neural network was discussed and theoretically analyzed, and it was pointed out that only using the network-groups could resolve the problem. On this foundation, library-similar knowledge-increasable neural network and its organizing algorithm were put forward, which could inherit availably the knowledge of the individual network in it and had the knowledge-increasing ability. Simulating results show that it is feasible and effective.
出处
《系统仿真学报》
CAS
CSCD
北大核心
2009年第1期100-103,共4页
Journal of System Simulation
基金
国家自然科学基金(60474014
60774046)
教育部高等学校博士学科点专项基金(20040151007)
交通部应用基础研究资助(200432922504)
关键词
人工智能
人工神经网络
知识可增殖神经网络
机器学习
artificial intelligence
artificial neural network
knowledge-increasable neural network
machine learning