机器学习

Based on LML theory frame and its algebraic model, geometric model and learning axiom systems, going for further study, this paper presented orbits generated algorithm of learning subspace in LML and applied it to corresponding examples, such as classify of human, chemical composition of wine and eight data sets including Soybean-Large, etc.

本文在李群机器学习的理论框架上，以李群机器学习的代数模型、几何模型、学习的公理系统为基础作进一步研究，给出了李群机器学习的学习子空间轨道生成算法，将该算法应用于人群分类，葡萄酒化学成分分类以及大豆等八个专用数据集的分类，取得了满意的结果。

Therefore, the main characteristics of this paper are:(1) Given orbits generated correlation theory of learning subspace which is the base of researching algorithm;(2) Advanced orbits generated Breadth first, Depth first and Heuristic algorithms, enrich and develop the basic content of LML further;(3) Given corresponding examples to validate the algorithms. However, all the work is tentive and much needs advanced research.

由此可以看出，本文的特色主要体现在以下几个方面：(1)给出了李群机器学习子空间轨道生成的相关理论，为研究学习子空间轨道生成算法奠定了基础；(2)提出了李群机器学习子空间广度优先轨道生成学习算法，深度优先轨道生成学习算法以及带有启发信息的轨道生成学习算法，丰富和发展了LML的基本内容；(3)对所提出的算法给出了相应的实例验证。

Aiming at the dual affects to classification results by each input data, this method has punish item be fuzzy, compensates weight to classification, reconstructs the optimization problem and its restrictions, reconstructs Langrage formula, and presents the theories deduction. This method is applied to the credit evaluating system of personal loan.

1引言支持向量机(Support Vector Machine，SVM)是20世纪90年代中期在统计学习理论(Statistical Learning Theory，SLT)的基础上提出的一种新的机器学习方法，它基于VC维理论和结构风险最小化原理，在很大程度上克服了传统机器学习中的维数灾难以及局部极小等问题[1，2]。

PAC is a statistical machine learning, machine learning and other methods of integration theory.

PAC是统计机器学习、集成机器学习等方法的理论基础。

This work tries to apply the learning process of the cognition structure defined in Cognitive Psychology to enhance or modify the development of AI, of which the learning models are almost based on trial and error style. However, this kind of learning style is definably given to the experience behavior of stimulus and response in Psychology. Thus, the relative AI models based on such style are design as an experience-adaptation system. For better ones, e.g. evolution-base algorithms, they belonged to the system with more powerful computing power to the dynamical environment. Even so, it was considered not only outside environment but also internal parameter tuning. As for the entire learning process, it has never been enhanced. That is, various original AI models are easily to be developed to their own close-form problem.

本研究企图以认知心理学之认知结构来修正自1956年以来人工智慧之发展，由於人工智慧长期局限於试误学习之低效率学习模式，然而试误学习於传统心理学定义中仅限於刺激与反应之经验行为而已，由此学习模式所建构之任何机器学习，均只能认定为经验之适应模式而已，而较进阶的种类，如演化式计算模型，也只是其能透过电脑强大的运算能力来达成所谓的动态环境下之演化式学习模式，其中演化之特色只是多考虑了外在环境的变化或内在参数的调整，而整个学习流程却没有进ㄧ步修正。

In the 1990s, a more complete theoretical system - the achievement of statistical learning theory and neural networks, and other than because of the emerging machine learning methods of the difficulties encountered some important, such as how to determine the issue of network structure, overfitting and underfitting, the local minimum points ect, makes the rapid development and improvement of SVM in resolving the small sample, non-linear and high-dimensional pattern recognition problems in the performance of many unique advantages and can function to promote the use of fitted function and other machines learning problems.

上世纪90年代，一个较完善的理论体系--统计学习理论( Statistical Learning Theory，简称SLT)的实现和由于神经网络等较新兴的机器学习方法的研究遇到一些重要的困难，比如如何确定网络结构的问题、过学习与欠学习问题、局部极小点问题等，使得SVM迅速发展和完善，在解决小样本、非线性及高维模式识别问题中表现出许多特有的优势，并能够推广应用到函数拟合等其他机器学习问题中。

The support vector machines based on ethe statistical learning theory is a new machine learning algorithm as the hotspots of machine learning research.

而基于统计学习理论的支持向量机方法是一种新的机器学习算法，已成为机器学习研究的热点。

The generalization ability of SVM is mainly depended on its parameters; therefore, unsuitable parameters can result in over-learning or lack-learning of SVM.

统计学习理论是专门研究小样本情况下机器学习的理论，其核心思想是通过控制学习机器的复杂度来控制学习机器的推广能力。

Inductive learning is one of the most important,the corest and the maturest branches in machine learning, but for using the knowledge acquired by learning and improving inductive learning algorithms, there are a lot of problems that are hard to solve using traditional methods.

1引言机器学习是当前人工智能的一个热门学科，归纳学习旨在从大量经验数据中归纳抽取一般的判定规则和模式，是机器学习最重要、最核心也是最成熟的一个分支。

Compared with other data mining platforms,such as MS SQL2005,Oracle 10g, WEKA obviously has advantages in supporting machine learning algorithms. WEKA has provided almost all kinds of machine learning algorithms and implemented in JAVA. But WEKA has a lot of disadvantages for data minging.

与其他主流挖掘平台相比，WEKA机器学习平台在支持机器学习模式识别算法方面具有明显优势，WEKA提供了目前所有主要机器学习算法的JAVA实现，但其直接用于数据挖掘还有很多不足。

The absorption and distribution of chromium were studied in ryeusing nutrient culture technique and pot experiment.

采用不同浓度K2CrO4(0，0.4，0.8和1.2 mmol/L)的Hoagland营养液处理黑麦幼苗，测定铬在黑麦体内的亚细胞分布、铬化学形态及不同部位的积累。

By analyzing theory foundation of mathematical morphology in the digital image processing, researching morphology arithmetic of the binary Image, discussing two basic forms for the least structure element: dilation and erosion.

通过分析数学形态学在图像中的理论基础，研究二值图像的形态分析算法，探讨最小结构元素的两种基本形态：膨胀和腐蚀；分析了数学形态学复杂算法的基本原理，把数学形态学的部分并行处理理念引入到家实际应用中。