【作者】 胡东红；

【导师】 李德华；

【作者基本信息】 华中科技大学 ， 模式识别与智能系统， 2004， 博士

【摘要】 钱学森院士倡导的思维科学是一门研究人的思维规律、历史发展和人工模拟的综合性科学技术大部门,它由基础科学、技术科学和应用技术三层次构成。由于大脑极其复杂,很难通过直接的方法了解大脑的思维机制,有许多问题仍然处于描述阶段。李德华教授采取同多学科工作者(科学哲学、心理学、科学史、方法论和各相关学科)合作的方式,从包含大量信息的发明创造案例入手(思维实践材料),分析了不同领域、类型的众多发明创造案例,从中抽取一般规律(在同中求异、在异中求同),总结出一种创造性思维模型——可能性构造空间理论 PCST。认为创造性思维从本质上说就是构造新的满足约束条件的可能性虚构元素及其空间,并在其中进行选择的反馈过程。并最先获得国家的支持,得到了国家重大基础研究“攀登计划”项目:认知科学前沿领域若干重大问题研究;国家“863 计划”智能计算机问题的基础研究项目:创造性思维模型及其实现技术研究、创造性思维中可能性构造空间理论研究;国家自然科学基金委员会主任项目:记忆与思维的计算模型研究等多个研究项目的资助。为了探讨思维科学的一般规律,为了探讨中医学的知识处理的规律,我们采用理论结合实际的方法,将创造性思维中的可能性构造空间理论与中医药的具体领域知识相结合。比较深入地探讨了思维和知识处理的一般规律。同时,作为国家 973 课题的子课题之一,本文按照 973 项目的要求,在可能性构造空间理论、中医药创新配方系统以及试验设计等多个方面进行了研究,主要贡献如下:1.提出了关于知识处理的潜在规则的设想:知识表达与推理的代价最小化原则。在分析中医学知识的基本知识表达方式,以及这种表达方式与背景知识的关系的基础上,提出了关于跨领域知识处理的潜规则的设想:知识表达与推理的代价最小化原则。在对中药研究非常有用的试验设计方法研究方面提出了均匀设计中的均匀性度量的三种新算法。2.提出了均匀性度量中的密集性偏差与稀疏性偏差。根据偏差的思想,以最理想的布点为标准,提出了基于最大最小空穴半径的均匀性度量的稀疏性度量和密集性度量准则及其计算方法。该方法的特点是计算量不大,不仅能较好地度量布点的均匀性以及布点在低维投影的均匀性,而且能指导如何调整布点使之尽可能与理想布点接近。最大空穴与最小空穴的半径与标准半径的比较,客观地从稀疏程度与密集程度两 I<WP=5>个方面描述了布点的均匀性. 而最大空穴与最小空穴的位置,又为如何调整布点,使之尽可能接近标准布点指明了调整方向. 3.提出了均匀性度量中的快速模板匹配算法。我们直观地发现,将空间延拓后, 研究点的数量 i 与内含 i 点的最大方体和最小方体的体积, 所得到的偏差符合偏差的定义。不仅可以消除原点的特殊地位, 而且可以给出偏差的点分布特性和空间分布特性。而偏差的点分布特性和空间分布特性可以清晰地表明布点的密度分布特点。对如何进一步调整布点, 使之最均匀, 具有重要意义。 4.提出了均匀性度量中的势函数模型。根据均匀性度量应满足的基本条件,根据物理学的势和力的概念,提出了均匀性度量的势函数模型。该模型较好地解决了均匀性度量在可计算性、布点的均匀性调整方法等方面的问题;而且具有均匀性度量所应该具有的许多优良性质,如旋转对称性、平移对称性、中心对称性等。最后给出了应用实例,并就低维投影的均匀性问题和势函数模型的改进作了讨论。 5.提出了中药单味药性状特征空间的概念。根据创造性思维模型中的可能性构造空间理论(PCST)和中药单味药的特性,提出了中药单味药性状特征空间及其基的概念,并分析了一个实例。为从中医的角度,较科学地描述自然药物特性及药效作用,提出了一种可能的途径。 6.提出了中医的四诊特征空间与辨证特征空间的概念。分析了中医知识的基本特点,根据创造性思维中的可能性构造空间理论和中医药创新配方系统的总体设计,提出了中医知识形式化的基本模式——中医的四诊特征空间与辨证特征空间。定义了空间元素的结构——参数与方法。把中医的诊断知识分解为具体的从四诊特征空间到辨证特征空间的网络连接关系。把中医组方的知识分解为四诊特征空间和辨证特征空间到中药特征空间的网络连接关系。 7.初步建立了中药复方创新配方的创造性思维过程。运用可能性构造空间理论(PCST)及其选择模型,初步建立了中药复方创新系统中的属性集和规则集,以及各种算子,给出了系统运行的一个简单的创造性思维的过程。 8.提出了基于 PCST 的中医药创新配方系统的组成框架。在分析中医药新药研制与发现过程中复杂的创造性思维过程特点及创造性思维中可能性构造空间理论模型的基础上,提出了建立一个旨在应用人工智能技术(Artificial Intelligence)、人工生命学(Artificial Life)理论、数据库技术、多媒体等技术的为中医药研究提供数据分析的中医药创新配方系统的组成框架。更多还原

【Abstract】 Thinking Science initiated by Academician Qian Xuesen is the synthetical sciencestudying the regular pattern, historic development and artificial imitation in human being’sthinking. It comprises three parts: basic science, engineering science and appliedtechnology. Creative thinking is the basic form of human beings’ discoveries and inventions.The purpose of studying creative thinking is to extend human’s ability. Professor Li Dehua has founded a model of creative thinking — PossibilityConstruction Space Theory (PCST). He holds that creative thinking is in essential areacting process, during which new satisfying possibly fictional elements & space areconstructed. Within the same process, choices are made among them. Professor Li Dehua’sresearch is the first to win the nation sponsors. His research is supported by the project “Onseveral weighty problems in the forward field of cognitive science”, which was one ofNational Key Project of fundamental research “Climbing Program”. It is also supported bythe projects “On creative thinking model and practical technology” and “On the calculationmodel of memory and thinking”, which are respectively fundamental research of IntelligentComputer, National “863 Project” and chairman’s Item of National Science Fundcommittee. For researching the general rule of thinking science and the general rule of knowledgemanipulation of the CTM (Chinese traditional medicine), we combined the theory with thepractice. Combining the PCST with the knowledge of the CTM, we have made someresearch on the thinking and knowledge processing. As a branch of the National “973 Subject”, this paper follows the requirements of thesubject. It researched on areas related to the Traditional Chinese Medicine experimentsdesigning and has made some contributions to the design of uniform branches ofexperiment designing. It also studied on the analyses of the knowledge structure of ChineseMedicine, provided our views on the internal reasons for the special expression of ChineseMedicine and proposed several thinking models related to Chinese Medicine. Thefollowing are our contributions: 1. The hypothesis about the knowledge processing is proposed: The cost III<WP=7>minimization of knowledge representation and inference. 2. New measurements of uniformity-compression discrepancy and sparsenessdiscrepancy are proposed. Based on the discrepancy criteria of uniformity measurementproposed by Wang Yuan and Fang K.T., standard radius of ideal distribution is defined. Fdistance and g distance between two points in m dimensional cubic space are defined, andso, maximum cavity and minimum cavity are defined. New measurements ofuniformity-compression discrepancy and sparseness discrepancy are proposed. The resultin two dimensions is calculated. With little calculation, our method can not only measurethe uniformity of distributions well and get a good uniformity in low dimensionalprojection, but also can give advices on how to adjust the distribution and let it be as nearto ideal distribution as possible. 3. The algorithms for getting the minimum and maximum templets for uniformmeasurement are proposed. Space is copied and repeated. Minimum templet andMaximum templet including i dots are defined. Also, the algorithms for getting the templetsare proposed. Our algorithm not only can solve the problems such as computability ofdiscrepancy, the special status of the origin and etc., but also can give the characters of thedots and the space distribution of the discrepancy. 4. The potential function model of uniformity measurement is proposed. The basiccharacteristics that any criterion of uniformity measurement should have are discussed.According to the model of physics potential and force, a new criterion of uniformitymeasurement, potential function model of uniformity measurement, is proposed. The modelnot only can solve the problems such as computability and t更多还原