基于支持向量机的模式识别摘要随着人工智能和机器学习学科的不断发展,传统的机器学习方法已经不能适应学科的快速发展。
而支持向量机(Support Vector Machine,SVM)则是根据统计学习理论提出的一种新型且有效的机器学习方法,它以结构风险最小化和VC 维理论为基础,适当的选择函数子集和决策函数,使学习机器的实际风险最小化,通过对有限的训练样本进行最小误差分类。
支持向量机能够较好的解决小样本、非线性、过学习和局部最小等实际问题,同时具有较强的推广能力。
支持向量机的样本训练问题实质是求解一个大的凸二次规划问题,从而所得到的解也是全局最优的,通常也是唯一的解。
本文以支持向量机理论为基础,对其在模式识别领域的应用进行系统的研究。
首先运用传统的增式支持向量机对历史数据分类,该分类结果表明对于较复杂的数据辨识时效果不佳。
然后运用改进后的增式支持向量机对历史数据进行分类,再利用支持向量机具有的分类优势对数据进行模式识别。
本文对传统增式支持向量机算法和改进增式支持向量机算法进行了仿真对比,仿真结果体现了改进增式支持向量机算法的优越性,改进增式支持向量机算法减少了训练样本集的样本数量,优化了时间复杂度和空间复杂度,提高了分类效率。
该方法应用于模式识别领域中能明显提高系统的准确率。
关键词:支持向量机;模式识别;多类分类;增式算法Pattern Recognition Based on Support Vector MachineAbstractWith the discipline of artificial intelligence and machine learning continues to evolve, traditional machine learning methods can not adapt to the rapid development of disciplines. The support vector machine (Support Vector Machine, SVM) is based on statistical learning theory a new and effective machine learning method, which to base on the structural risk minimization and the VC dimension theory, a function subset of appropriate choice and decision-making function of appropriate choice, the learning machine to minimize the actual risk, through the limited training samples for minimum error classification. SVM can solve the small sample, nonlinear, over learning and local minimum practical issues, but also it has a strong outreach capacity. Sample training problems of Support Vector Machines to solve really a large convex quadratic programming problems, and to the global optimal solution is also obtained, usually the only solution.This paper based on support vector machine theory, its application in the field of pattern recognition system. First, by using the traditional incremental support vector machine classification of historical data, the classification results show that the data for the identification of more complex when the results are poor. And then improved by the use of incremental Support Vector Machines to classify the historical data, and then use the classification of Support Vector Machine has advantages for data pattern recognition.This type of traditional incremental Support Vector Machine and improved incremental Support Vector Machine algorithm was simulated comparison, simulation results demonstrate the improved incremental Support Vector Machine algorithm by superiority, improved incremental Support Vector Machine algorithm reduces the set of training samples number of samples,and to optimize the time complexity and space complexity, improving the classification efficiency. The method is applied to pattern recognition can significantly improve the accuracy of the system.Key words: Support Vector Machine; Pattern Recognition; Multi-class Classification; Incremental Algorithm目录论文总页数:37页第一章引言.................................................错误!未定义书签。
1.1课题背景及意义........................................错误!未定义书签。
1.2支持向量机理论的发展...................................错误!未定义书签。
1.3支持向量机在各个领域的应用.............................错误!未定义书签。
1.4本课题意义及内容安排...................................错误!未定义书签。
第二章支持向量机的基本原理................................错误!未定义书签。
2.1统计学习理论..........................................错误!未定义书签。
2.1.1机器学习问题描述...............................错误!未定义书签。
2.1.2统计学习理论的发展.............................错误!未定义书签。
2.1.3 VC维理论......................................错误!未定义书签。
2.1.4 推广性的界....................................错误!未定义书签。
2.1.5结果风险最小化原则.............................错误!未定义书签。
2.2支持向量机(SVM)理论....................................错误!未定义书签。
2.2.1最优分类面.....................................错误!未定义书签。
2.2.2标准支持向量机.................................错误!未定义书签。
2.3支持向量机的主要研究方法...............................错误!未定义书签。
2.3.1支持向量机多类分类方法.........................错误!未定义书签。
2.3.2解决支持向量机的二次规划问题...................错误!未定义书签。
2.2.3核函数的选择及其参数优化.......................错误!未定义书签。
2.4本章小结..............................................错误!未定义书签。
第三章基于增式支持向量机的模式识别.........................错误!未定义书签。
3.1传统增式SVM训练算法.................................错误!未定义书签。
3.2改进增式SVM训练算法..................................错误!未定义书签。
3.2.1改进算法的基本思路及KKT条件...................错误!未定义书签。
3.2.2 改进增式训练算法的步骤........................错误!未定义书签。
3.2.3 仿真实验......................................错误!未定义书签。
3.4本章小结..............................................错误!未定义书签。
第四章改进SVM算法在模式识别领域的应用....................错误!未定义书签。
4.1模式识别简介.........................................错误!未定义书签。
4.2舰船目标识别..........................................错误!未定义书签。