自主机器人人工进化方法研究

在介绍人工进化研究状况的基础上 ,着重阐述了进化机器人学的关键问题 :基本概念、进化技术、进化计算和实验研究方法等 ,探索了进化机器人学未来的研究趋势及所面临的挑战 .     

用遗传算法求解课程表问题

课程表问题是NP完全类问题.近些年来人们尝试着用进化算法求此问题.本文根据大学编排课表的特点设计了一种全新的编码和适应值函数,并应用遗传算法求解.试验说明了该方法的可行性和有效性.     

求代数方程组数值解的智能新方法

本文针对解代数方程组计算复杂和非线性方程组的解难以得到问题，提出了一种适合于不同类型方程组的通用算法．模拟生物进化过程，利用仅以变异作为唯一基因操作的EP方法来求方程组的最优解或次最优解．首先建立智能化的通用方程组，再利用改进的EP方法(在自适应方法中引入小生境思想)来求解方程组．算法既简单又具有通用性，最后举例说明本方法的有效性．     

Starch-binding domains in the post-genome era

Starch belongs to the most abundant biopolymers on Earth. As a source of energy, starch is degraded by a large number of various amylolytic enzymes. However, only about 10% of them are capable of binding and degrading raw starch. These enzymes usually possess a distinct sequence-structural module, the so-called starchbinding domain (SBD). In general, all carbohydrate-binding modules (CBMs) have been classified into the CBM families. In this sequence-based classification the individual types of SBDs have been placed into seven CBM families: CBM20, CBM21, CBM25, CBM26, CBM34, CBM41 and CBM45. The family CBM20, known also as a classical C-terminal SBD of microbial amylases, is the most thoroughly studied. The three-dimensional structures have already been determined by X-ray crystallography or nuclear magnetic resonance for SBDs from five CBM families (20, 25, 26, 34 and 41), and the structure of the CBM21 has been modelled. Despite differences among the amino acid sequences, the fold of a distorted β-barrel seems to be conserved together with a similar way of substrate binding (mainly stacking interactions between aromatic residues and glucose rings). SBDs have recently been discovered in many non-amylolytic proteins. These may, for example, have regulatory functions in starch metabolism in plants or glycogen metabolism in mammals. SBDs have also found practical uses. Received 25 May 2006; received after revision 26 June 2006; accepted 3 August 2006     

Finding global minima with a new dynamical evolutionary algorithm

A new dynamical evolutionary algorithm (DEA) based on the theory of statistical mechanics is presented. This algorithm is very different from the traditional evolutionary algorithm and the two novel features are the unique of selecting strategy and the determination of individuals that are selected to crossover and mutate. We use DEA to solve a lot of global optimization problems that are nonlinear, multimodal and multidimensional and obtain satisfactory results. Foundation item: Supported by the National Natural Science Foundation of China (No. 60133010, NO. 60073043 and No. 700/1042) Biography: Zou Xiu-fen(1996-), female, Ph. D candidate, Associate professor, research direction: evolutionary computing, parallel computing.     

多维函数优化的遗传算法研究

本文研究了求解多变量函数优化问题的遗传算法,在此算法中采用了十进制浮点数基因表示方法,并相应地提出了一种叠加零均值Gauss随机扰动的变异方法,研究表明,对于满足组件假说的多维函数优化问题,这种遗传算法具有较高的搜索效率．     

三种国外冷杉的核型研究

首次报道了３种国外冷杉的核型。ＡｂｉｅｓｃｅｐｈａｌｏｎｉｃａＬｏｕｄ。有９对中部着丝粒和３对近中着丝粒染色体,Ａ．ｎｕｍｉｄｉｃａＤｅＬａｎｎｏｎｅｘＣａｒｒ．文中还讨论了３种冷杉及其氯不组的进化地位。     

电力系统潮流计算的模拟进化方法

将模拟进化方法引入电力系统潮流计算。与传统方法比较结果表明,模拟进化方法中的进化策略法可以有效地解决电力系统病态潮流问题。文中还对几种模拟进化方法进行了比较,结果表明：进化策略法用地潮流计算比较成功;基因算法和进化规划法由于方法本身的缺陷,不太适合于潮流计算。     

用进化规划思想优化径向基函数神经网络结构的均衡器

针对训练径向基函数(RBF)神经网络均衡器的随机梯度算法(SG)中,神经网络的结构是指定的并且所用训练样本较长的问题,引入进化规划思想,用进化规划方法确定径向基函数神经网络的结构,用基于最小均方(LMS)误差准则的自适应算法调整神经元到输出端的连接权重.蒙特卡洛仿真表明,用这种方法确定的均衡器可以达到与SG算法相同的性能,而所用训练样本很少,网络结构不需要事先指定.     

认知进化观:认知科学研究范式的新取向

认知科学在其发展过程中经历了几次研究范式的转换，认知进化观的兴起是认知科学各学科多层次整合的表现。在心理学领域，进化心理学采用功能分析的方法分析适应性行为下的心理机制和特征。在人工智能领域，采用进化计算和遗传算法探讨人工生命的适应性行为。在神经科学领域，是对神经网络的自组织机制和神经系统的可塑性进行考察。认知进化观的缺陷是自然主义的计算与进化，忽视了人的历史文化性。认知进化观作为一种新方法论，为智能研究带来了生机。