基于神经信息的汉语元认知层级建模研究

项目名称： 基于神经信息的汉语元认知层级建模研究

项目编号： No.61073056

项目类型： 面上项目

立项/批准年度： 2011

项目学科： 轻工业、手工业

项目作者： 刘洪波

作者单位： 大连海事大学

项目金额： 12万元

中文摘要： 汉语微文广义，是稳定收敛的二维信息，生动而且高效。汉语元认知研究为开启智能之门提供了机遇。本项目在神经信息学的基础上研究汉语元认知建模，层次性地分析人脑在汉语认知中的状态及其迁移过程，解析汉语认知和元认知，研究人脑网络动态地组织和重塑特性，经多知识粗糙集理论对稀疏节点进行约简，采用贝叶斯网络分析网络节点间的因果关系，基于多代理和网络构建层级模型。该模型的顶层在较大尺度上具有网络结构，采用自上而下的指导性学习机制，底层在较小尺度上是相关联的分布式多代理系统，采用自下而上的选择性学习机制。两个层级分别使用与外显和内隐加工相适应的符号表征和分布式表征。系统中的代理个体具有生命周期，在学习和内省中创建，在其生命周期内以黄金分割概率被新陈代谢，表现出涌现性。该项研究具有中国特色和优势，将会为设计相应的人工认知系统提供新的计算模型和方法。

中文关键词： 神经信息学；汉语认知；神经网络；多代理系统

英文摘要： Chinese language could provide refined representations, which is convergent and provides stable two-dimension information. The research on language cognition would provide us the opportunities to uncover the complex relationship between mind and consciousness. This project focuses on meta-cognitive modeling based on Neuroinformatics Research. We make an attempt to formulate a two-layer model for Chinese cognition and meta-cognition. There is a larger scale network structure in the top layer of the model, using top-down guidance of learning mechanism. And in its bottom, there is a distributed multi-agent system at smaller scales, using a bottom-up selective learning mechanism. The two levels are used with the explicit and implicit processing corresponding symbolic representation and distributed representation. We believe that the works would help us to formulate the relationship between the time series and understand the dynamic processes of Chinese cognition, which would provide tools to draw the dynamic Chinese cognition networks. It is investigated with the help of the theory of complex networks. Some numerical simulation and experiments are made for testing the anti-dilapidated ability of the brain functional networks. This research would be a significant step to understand the functioning of the human brain.

英文关键词： Neuroinformatics; Chinese cognition; Nerural networks; Multi-agent system