多智能体系统中强化学习模型的改进及应用

多Agent系统中的强化学习是近年发展起来的一种新的人工智能方法,是以单Agent强化学习Q-learning算法为基础的一种学习模型,由于现有的强化学习模型还存在着结构信度分配困难、学习速度慢等缺陷,这些缺陷大大限制了多Agent强化学习模型的应用范围,本文对多Agent强化学习模型进行了系统的研究,分析了多Agent理论中强化学习面临的任务,指出在多Agent系统顺序型任务中遇到的时间信度分配难题及多Agent系统Agent间＂状态-动作＂对Q值估计的互通问题,对此问题提出了初步的解决办法,并在此基础建立了一个改进的多Agent强化学习模型,而且把该模型应用于电磁辐射源识别工作中。     

分层强化学习原理研究

首先介绍了强化学习基本原理,分析了马尔科夫决策过程与半马尔科夫决策过程的理论基础及其在强化学习中的应用,其次阐述了分层强化学习中分层与抽象的思想,分析了HAM、Options与MaxQ等方法,并从分层与抽象角度进行了比较,最后指出了分层强化学习的研究发展方向。     

电子商务领域移动代理的安全技术研究

分析了移动代理在电子商务环境中存在的安全问题,提出了一种移动代理在电子商务环境下的安全模型。该模型可以保护移动Agent在迁移过程中的安全,同时可以通过事后的分析认证找出攻击者。并引入访问控制策略和信用机制来降低移动Agent迁移过程被攻击的危险。通过分析论证,此安全模型能满足保护移动Agent迁移过程中安全性需求。     

一种基于多Agent协作的下载模型

提出了一种基于多Agent协作的下载模型,该模型根据协作规则,将下载任务智能、动态地分配给网络上多个Agent,利用多Agent协作快速、高效、灵活地完成下载任务。Agent的知识是在多Agent的协作过程中,通过学习自动产生、完善和优化的。     

Mobile Agent技术在局域网中的信息检索的实现

论述了一个应用于局域网中的基于Mobile Agent技术的信息检索系统的模型结构和关键技术的实现方法。整个 模型是建立在已有网络协议的基础上并且采用了分层结构,共分为四层:网络层、平台层、代理层、应用层。关键技术涉及 Agent的基本结构、Agent的管理和通信、Agent的回调生成算法和安全问题。     

Agent在网格资源监控系统中的应用

掌握网格资源状态和负载变化是网格调度和应用性能分析与预测的基础。通过提出一个基于Agent和LDAP协议的GridFound系统,重点进行LDAP目录设计,对系统功能进行阐述。本系统利用移动Agent特点,在异构网络及分布式计算环境中自主和自动地迁移、携带信息和寻找信息资源,进行信息处理,完成任务请求。     

智能代理技术在健美操教学中的应用

智能代理技术(Intelligent Agent)是20世纪90年代以来人工智能领域重要的研究内容。根据健美操教学的特点,通过分析智能代理技术的功能和特点,提出了将智能代理技术应用在健美操教学中的学习模型。     

基于感知行动的一种双向规划算法研究

规划最主要的内容是对Agent行动进行推理,但Agent对真实世界环境的知识是不完备的,因此这种情况下的规划,需要感知行动以扩充Agent的知识状态,带有感知行动的规划,其结果不再是一个简单的行动序列,而是带有感知行动及其条件说明的条件规划。本文根据Agent知识状态空间的搜索方向,讨论了正向规划、反向规划,最后提出了一个能够处理感知行动和条件规划的双向规划算法BCPS(Bi-direction Conditional Planning with Sensing actions)。     

基于感知行动的一种双向规划算法研究

规划最主要的内容是对Agent行动进行推理,但Agent对真实世界环境的知识是不完备的,因此这种情况下的规划,需要感知行动以扩充Agent的知识状态,带有感知行动的规划,其结果不再是一个简单的行动序列,而是带有感知行动及其条件说明的条件规划。本文根据Agent知识状态空间的搜索方向,讨论了正向规划、反向规划,最后提出了一个能够处理感知行动和条件规划的双向规划算法BC PS(Bi-direction Conditional Planning with Sensingactions)。     

促进高中学生政治知识迁移的策略思考

学习过程往往伴随着迁移现象的发生,学习的强化离不开迁移。在学校无法传递一切知识给学生的情况下,最好的办法是培养学生自己获取知识的能力,实现教即是为了迁移的目的。而思想政治在学生成长方面的作用不言而喻,其要发挥好该学科在基础教育中的感化功能。要想实现高中政治学习中的知识迁移,离不开迁移策略的掌握和应用,高中政治学习活动中学习迁移能力的提高,是实现教育教学工作的高效和学生驾驭知识的重要保证。     

The Effect of a Mastery Learning Technique on the Performance of a Transfer of Training Task

This study evaluated the effect of using the mastery learning technique of self‐directed feedback, reinforcement, and remediation of knowledge on the performance of a work‐related task. The mastery learning intervention was conducted via a workbook which provided feedback to participants on their knowledge attainment after instruction, yet before the evaluation of the transfer task. The study utilized a randomized subjects, post‐test‐only control group design with 130 participants. The hypothesis that mastery learning would have a positive effect on transfer of knowledge from the classroom to a work‐related task was supported. Implications for instructional design and suggestions for future research are discussed.     

基于技术实现知识有效迁移的方式探讨

迁移是有效学习的基本特征之一,也是教育要达到的目标之一。只有真正实现了知识的迁移,才能利用所学知识来解决各种问题。通过技术可以有效促进知识的迁移,即通过技术提供模拟训练,引导知识迁移;利用技术支撑学习活动,启迪知识迁移;利用技术创设学习环境,促进知识迁移;基于技术创建学习共同体,协同知识迁移。     

学生所知道的 教师所知道的——理论、工具和结果

评估——无论采用何种形式——不得不建立在对不同水平和质量的理论思考基础之上。其中至关重要的是测验内容领域要依据教与学的过程来构建。其次,任务是评估相关内容结构的核心工具。因此,在设计对数学的教育评估时,按照合理的标识和特征对任务进行分类就成为一个息息相关的问题,也是一个巨大的挑战。本文重点研究PISA(国际学生评估项目的缩写,是以数学素养为目标的国际比较研究,包括在国家层面德国所做的许多调整)和COACTIV(德国教师数学知识和数学教学知识的一个代表性研究)。两者都旨在实现区分性评估(differentiated assessment),都考虑了评价内容和任务分析的不同方面。区分性评估结果使得我们能够获取有用的信息,进一步推动课堂教与学的发展。这些信息主要表现为两个方面:评估学生不同类型的知识,这些知识类型直接指向课堂教学与学习中存在的各种问题;探索教师知识对学生学习进展的影响,表明哪种教学行为是有效的。     

The ontologies of complexity and learning about complex systems

This paper discusses a study of students learning core conceptual perspectives from recent scientific research on complexity using a hypermedia learning environment in which different types of scaffolding were provided. Three comparison groups used a hypermedia system with agent-based models and scaffolds for problem-based learning activities that varied in terms of the types of text based scaffolds that were provided related to a set of complex systems concepts. Although significant declarative knowledge gains were found for the main experimental treatment in which the students received the most scaffolding, there were no significant differences amongst the three groups in terms of the more cognitively demanding performance on problem solving tasks. However, it was found across all groups that the students who enriched their ontologies about how complex systems function performed at a significantly higher level on transfer problem solving tasks in the posttest. It is proposed that the combination of interactive representational scaffolds associated with NetLogo agent-based models in complex systems cases and problem solving scaffolding allowed participants to abstract ontological dimensions about how systems of this type function that, in turn, was associated with the higher performance on the problem solving transfer tasks. Theoretical and design implications for learning about complex systems are discussed.     

基于支持向量机和转换的错误驱动学习方法的组块识别

支持向量机在高维特征空间的输入数据上具有较高的泛化性能,能够独立于小范围内的数据维数计算.基于转换的学习方法能自动融合不同类型的知识,所得到的模板可以显示一些语言知识,这些语言知识对于语言学及其他相关研究有重要意义.利用支持向量机和基于转换的错误驱动学习相结合,能够达到较为满意的组块识别效果.     

The Effects of Study Tasks in a Computer-Based Chemistry Learning Environment

The present study examines the effects of different study tasks on the acquisition of knowledge about acids and bases in a computer-based learning environment. Three different task formats were selected to create three treatment conditions: learning with gap-fill and matching tasks, learning with multiple-choice tasks, and learning only from text and figures without any additional tasks. Participants were 196 ninth-grade students who learned with a self-developed multimedia program in a pretest–posttest control group design. Research results reveal that gap-fill and matching tasks were most effective in promoting knowledge acquisition, followed by multiple-choice tasks, and no tasks at all. The findings are in line with previous research on this topic. The effects can possibly be explained by the generation–recognition model, which predicts that gap-fill and matching tasks trigger more encompassing learning processes than multiple-choice tasks. It is concluded that instructional designers should incorporate more challenging study tasks for enhancing the effectiveness of computer-based learning environments.     

建筑工程钢筋翻样技能任务教学法探析

建筑工程钢筋翻样是一项重要而实用的专业技能之一。在钢筋翻样技能教学过程中,应用任务教学法,将学生的学习活动与任务或问题相结合,让学生带着真实的任务去学习,不仅有助于提升学生钢筋翻样的专业岗位职业能力,同时有助于培养学生终身发展所需要的自主学习能力,以及将来走向工作岗位所应具备的沟通交流、团队协作等方面社会能力。     

从补缺假说看语言测试真实性的意义

语言测试的真实性是指，受试者在测试中使用目标语完成测试任务与其在现实生活中完成任务的相似程度。补缺假说强调在学习外语时，要同时学习外语语境知识。二者都强调语境的作用。因此，本文认为，既然补缺假说要求我们无论是外语的教还是学都必须在语境中进行，而测试又对教学有巨大的反拨作用，我们在设计测试时，就应该遵循语言测试的真实性原则，把测试置于真实的语境中，以促进教师和学生对外语语境知识的教学的重视。     

学习过程中愉悦心理机制的构建

有效的学习过程应体现为，学生形成稳固的、愉悦的学习心理，在轻松、愉快的氛围中掌握知识、技能，完成学习任务。这需要教师对学生进行正确的学习心理指导，加强情感教育，培养学习兴趣，使学生正确面对学习挫折，在学习过程中有成功的快乐体验。     

Enhancing learning and problem solving skills: orienting and self-judging, two powerful and trainable learning tools

In this study we report on orienting and self-judging as study and problem-solving activities or learning tools. We examine their trainability and their effect on academic performance. These questions are part of a research project aimed at improving metacognitive knowledge as well as affective, conative and regulation skills. A design experiment with 141 freshmen in business economics was set up. The experimental treatment consisted of an integrated set of instructional conditions which were operationalized in a series of sessions and practice and transfer tasks. The results show that after the intervention students of the experimental group had more knowledge about orienting and self-judging than the students of both control groups and they also oriented themselves better and were more prone to self-judging when starting a new course. Both metaknowledge and transfer behavior were positively related to academic performance.