电子学报 ›› 2018, Vol. 46 ›› Issue (8): 2011-2019.DOI: 10.3969/j.issn.0372-2112.2018.08.029

• 学术论文 • 上一篇    下一篇

基于SA-ANN的认知机制建模与识别优化算法

陈树婷1, 谭大鹏2   

  1. 1. 杭州医学院基础医学部, 浙江杭州 310053;
    2. 浙江工业大学机械工程学院, 浙江杭州 310032
  • 收稿日期:2017-03-16 修回日期:2017-11-13 出版日期:2018-08-25
    • 通讯作者:
    • 谭大鹏
    • 作者简介:
    • 陈树婷 女,1981生于山东莱芜.现为杭州医学院基础医学部教师.主要研究方向为人工智能、模式识别.E-mail:shutinren@163.com.
    • 基金资助:
    • 国家自然科学基金 (No.51775501); 浙江省杰出青年科学基金 (No.LR16050001); 浙江省医药卫生科技计划 (No.2015KYA067); 浙江省教育科学规划项目 (No.2017SCG386)

SA-ANN Based Cognition Mechanism Modeling and the Improved Recognition Algorithm

CHEN Shu-ting1, TAN Da-peng2   

  1. 1. Department of Basic Medicine, Hangzhou Medical College, Hangzhou, Zhejiang 310053, China;
    2. College of Mechanical Engineering, Zhejiang University of Technology, Hangzhou, Zhejiang 310032, China
  • Received:2017-03-16 Revised:2017-11-13 Online:2018-08-25 Published:2018-08-25

摘要: 人脑认知过程机制建模是人工智能研究领域的重要方向,当前基于统计模板分析与反向传播神经网络(BP-ANN)的认知方法在聚类计算与知识理解方面存在不足.针对上述问题,提出了一种基于模拟退火神经网络(SA-ANN)的认知过程机制建模方法.对人脑认知物理过程及其基本特征进行了分析,建立了面向认知过程的SA-ANN推理模型.提出了一种改进的模拟退火神经网络(ISA-ANN)识别优化算法,对认知过程信息特征提取、知识学习训练等关键环节进行了模拟研究.设计了认知过程机制算例,开发了相应的原型软件系统,对理论结果进行了验证.结果证明,该方法具有较好的聚类性能,可以针对具体测试对象进行准确识别,能够得到相对精确的认知演化规律.

关键词: 认知过程, 机制建模, 信息加工, 人工神经网络, 模拟退火

Abstract: Human cognition mechanism modeling was an important research direction of artificial intelligence area.Current modeling methods based statistical modes or back propagation artificial neural network (BP-ANN) have the problems of clustering computation and knowledge understanding.Concerning the issue,a mechanism modeling method for cognition process was proposed based on simulated annealing artificial neural network (SA-ANN).The cognition physical process and its fundamental characteristics were analyzed,and a SA-ANN inference model oriented to cognition process was set up.An improved simulated annealing artificial neural network (ISA-ANN) processing algorithm was put forward,and the critical factors of information character extraction and knowledge clustering for cognition process were simulated.Numerical instances for human cognition were provided,and a prototype prototype software system was developed to verify the theoretical results.Experimental results prove that the proposed method is with better clustering performance,can correctly recognize the testing object selected,and can reveal the evolution regulars of psychological cognition process.

Key words: cognition process, mechanism modeling, information processing, artificial neural network, simulated annealing

中图分类号: