问题之间共性关系的意识水平对平面几何解题迁移的影响

Transfer of learning is one of the major concepts in educational psychology. As cognitive psychology develops, many researchers have found that transfer plays an important part in problem solving, and the awareness of the similarity of related problems is important in transfer. So they become more interested in researching the problem of transfer. But in the literature of transfer research, it has been found that many researchers do not hold identical conclusions about the influence of awareness of related problems during problem solving transfer. This dissertation is written on the basic of much of sub-research work, such as looking up literature concerning transfer of problem solving research, comparing the results of research work done recently and experimental researches. The author of this dissertation takes middle school students as subjects, geometry as materials, and adopts factorial design in his experiments. The influence of awareness of related problems on problem solving transfer is examined from three dimensions which are the degree of difficulty of transfer problems, the level of awareness of related problems and the characteristics of subjects themselves. Five conclusions have been made after the experimental research: (1) During the process of geometry problem solving, the level of awareness of related problems is one of the major factors that influence the effect of problem solving transfer. (2) Either more difficult or more easy of the transfer problems will hinder the influence of awareness of related problems during problem solving transfer, and the degree of difficulty of the transfer problems have interactions with the level of awareness of related problems in affecting transfer. (3) During geometry problems solving transfer, the level of awareness of related problems has interactions with the degree of student achievement. Compared with the students who have lower achievement, the influence of the level of the awareness is bigger in the students who have higher achievement. (4) There is positive correlation between geometry achievement and reasoning ability of the middle school students. The student who has higher reasoning ability has higher geometry achievement, while the level of awareness is raised, the transfer achievement of both can be raised significantly. (5) There is positive correlation between geometry achievement and cognitive style of the middle school students. The student who has independent field tendency of cognitive style has higher geometry achievement, while the level of awareness is raised, the transfer achievement of both can be raised significantly. At the end of the dissertation, the researcher offers two proposals concerning Geometry teaching on the basis of the research findings.

跨数据域迁移学习及相关问题的研究

在过去的二十年中，数据挖掘和机器学习受到了越来越多的关注。 这很大程度上是因为在互联网时代信息传播和积累的速度越来越快， 人工处理数据越来越困难，智能化及自动化的数据处理能力成为迫切的需求。 为此人们设计了很多学习算法，希望计算机能具有人类的学习能力，即只要训练一次，就 可以自动处理数据。 尽管这种学习能力已经在很多成功的应用中得到了验证，但它建立在一个重要的假设基础上，即训练数据与目标数据的一致性。 这意味着：根据训练数据得到的模型只适用于具有同样分布的目标数据。如果需要完成一个新的任务， 即使是与原任务非常相近的任务，原来训练好的模型也可能会失效。但是如果重新提供训练数据必将付出很高的成本。 因为两个任务之间存在的相似性，在新任务的训练过程中彻底丢弃原有的训练数据也是非常不合理的。 考虑到数据来源的差异性和训练数据的时效性在实际应用中普遍存在，有必要寻找更有效的解决途径。 迁移学习的提出正是为了解决上面的问题。传统的学习过程实际上是实现了从人到机器的知识迁移。 迁移学习则是研究从一个学习任务到另一个学习任务的知识迁移，以提高知识利用的效率。 这样的知识迁移将在缺乏训练数据和训练数据时效较短的情况下 大大降低学习的成本并提高学习的效率和自动化程度。 本文从跨数据域迁移学习入手，研究无监督迁移学习技术，以及在数据流环境下的有监督迁移学习技术， 在以下三个方面做出了创新性贡献： 在迁移学习中首次提出利用最大间隔方法在没有目标数据域的训练数据的情况下完成分类任务。 提出了两种算法，以迭代优化技术为基础，分别在函数层以及参数层实现了辅助任务到目标任务的知识迁移。 在多个公开的数据集中的实验表明，两种算法的分类准确率均优于现有的迁移学习算法。 在数据流分类任务中，针对概念漂移问题首次提出对概念漂移进行建模，来设计一种 可以自动适应数据分布变化的动态分类器。作为一种新的分类框架，可用于logistic regression和SVM等诸多分类模型。在实验中表明，所提出的算法有效避免了传统滑动 窗口方法导致的数据过拟合，实现了较高的分类准确率。 提出在具有多个节点的传感器网络中进行异常检测的新方法。利用主成分分析对数据空间进行变换，并根据能量阈值 对数据空间进行划分，构建异常子空间， 根据数据在异常子空间上的投影来检测异常数据点。基于数据点在异常子空间上的投影信息还可以进一步对异常来源 进行定位，并度量异常的大小。在实验中所提出的方法展现了较强的异常检测能力。

混沌粒子群混合算法对微型永磁电机的优化设计

在电机的设计中,常常需要通过优化设计得到合理的电机结构尺寸和参数.电机的设计问题实质上是一种带约束的复杂的非线性连续函数优化问题.要得到一个满意的优化结果不仅要求算法具有较高的精度,而且要有快的收敛速度.提出一种新的混合算法对永磁电机的尺寸和整体结构进行优化设计.将混沌算法和粒子群算法相结合,以微型永磁电机为例,对槽形等多个变量进行优化,结果证明了算法的有效性和快速性,适合于同类问题求解.