Incremental support vector machine algorithm based on multi-kernel learning

A new incremental support vector machine （SVM） algorithm is proposed which is based on multiple kernel learning. Through introducing multiple kernel learning into the SVM incremental learning, large scale data set learning problem can be solved effectively. Furthermore, different punishments are adopted in allusion to the training subset and the acquired support vectors, which may help to improve the performance of SVM. Simulation results indicate that the proposed algorithm can not only solve the model selection problem in SVM incremental learning, but also improve the classification or prediction precision.

Turbopump Condition Monitoring Using Incremental Clustering and One-class Support Vector Machine

Turbopump condition monitoring is a significant approach to ensure the safety of liquid rocket engine （LRE）.Because of lack of fault samples,a monitoring system cannot be trained on all possible condition patterns.Thus it is important to differentiate abnormal or unknown patterns from normal pattern with novelty detection methods.One-class support vector machine （OCSVM） that has been commonly used for novelty detection cannot deal well with large scale samples.In order to model the normal pattern of the turbopump with OCSVM and so as to monitor the condition of the turbopump,a monitoring method that integrates OCSVM with incremental clustering is presented.In this method,the incremental clustering is used for sample reduction by extracting representative vectors from a large training set.The representative vectors are supposed to distribute uniformly in the object region and fulfill the region.And training OCSVM on these representative vectors yields a novelty detector.By applying this method to the analysis of the turbopump＇s historical test data,it shows that the incremental clustering algorithm can extract 91 representative points from more than 36 000 training vectors,and the OCSVM detector trained on these 91 representative points can recognize spikes in vibration signals caused by different abnormal events such as vane shedding,rub-impact and sensor faults.This monitoring method does not need fault samples during training as classical recognition methods.The method resolves the learning problem of large samples and is an alternative method for condition monitoring of the LRE turbopump.

Fault-Diagnosis Method Based on Support Vector Machine and Artificial Immune for Batch Process

A new fault-diagnosis method to be used in batch processes based on multi-phase regression is presented to overcome the difficulty arising in the processes due to non-uniform sample data in each phase.Support vector machine is first used for phase identification,and for each phase,improved artificial immune network is developed to analyze and recognize fault patterns.A new cell elimination role is proposed to enhance the incremental clustering capability of the immune network.The proposed method has been applied to glutamic acid fermentation,comparison results have indicated that the proposed approach can better classify fault samples and yield higher diagnosis precision.

Using the Support Vector Machine Algorithm to Predict β-Turn Types in Proteins

The structure and function of proteins are closely related, and protein structure decides its function, therefore protein structure prediction is quite important.β-turns are important components of protein secondary structure. So development of an accurate prediction method ofβ-turn types is very necessary. In this paper, we used the composite vector with position conservation scoring function, increment of diversity and predictive secondary structure information as the input parameter of support vector machine algorithm for predicting theβ-turn types in the database of 426 protein chains, obtained the overall prediction accuracy of 95.6%, 97.8%, 97.0%, 98.9%, 99.2%, 91.8%, 99.4% and 83.9% with the Matthews Correlation Coefficient values of 0.74, 0.68, 0.20, 0.49, 0.23, 0.47, 0.49 and 0.53 for types I, II, VIII, I’, II’, IV, VI and nonturn respectively, which is better than other prediction.

The Comparison between Random Forest and Support Vector Machine Algorithm for Predicting β-Hairpin Motifs in Proteins

Based on the research of predictingβ-hairpin motifs in proteins, we apply Random Forest and Support Vector Machine algorithm to predictβ-hairpin motifs in ArchDB40 dataset. The motifs with the loop length of 2 to 8 amino acid residues are extracted as research object and thefixed-length pattern of 12 amino acids are selected. When using the same characteristic parameters and the same test method, Random Forest algorithm is more effective than Support Vector Machine. In addition, because of Random Forest algorithm doesn’t produce overfitting phenomenon while the dimension of characteristic parameters is higher, we use Random Forest based on higher dimension characteristic parameters to predictβ-hairpin motifs. The better prediction results are obtained;the overall accuracy and Matthew’s correlation coefficient of 5-fold cross-validation achieve 83.3% and 0.59, respectively.

Improved adaptive pruning algorithm for least squares support vector regression

As the solutions of the least squares support vector regression machine （LS-SVRM） are not sparse, it leads to slow prediction speed and limits its applications. The defects of the ex- isting adaptive pruning algorithm for LS-SVRM are that the training speed is slow, and the generalization performance is not satis- factory, especially for large scale problems. Hence an improved algorithm is proposed. In order to accelerate the training speed, the pruned data point and fast leave-one-out error are employed to validate the temporary model obtained after decremental learning. The novel objective function in the termination condition which in- volves the whole constraints generated by all training data points and three pruning strategies are employed to improve the generali- zation performance. The effectiveness of the proposed algorithm is tested on six benchmark datasets. The sparse LS-SVRM model has a faster training speed and better generalization performance.

Support vector machine incremental learning triggered by wrongly predicted samples

According to the classic Karush-Kuhn-Tucker(KKT)theorem,at every step of incremental support vector machine(SVM)learning,the newly adding sample which violates the KKT conditions will be a new support vector(SV)and migrate the old samples between SV set and non-support vector(NSV)set,and at the same time the learning model should be updated based on the SVs.However,it is not exactly clear at this moment that which of the old samples would change between SVs and NSVs.Additionally,the learning model will be unnecessarily updated,which will not greatly increase its accuracy but decrease the training speed.Therefore,how to choose the new SVs from old sets during the incremental stages and when to process incremental steps will greatly influence the accuracy and efficiency of incremental SVM learning.In this work,a new algorithm is proposed to select candidate SVs and use the wrongly predicted sample to trigger the incremental processing simultaneously.Experimental results show that the proposed algorithm can achieve good performance with high efficiency,high speed and good accuracy.

Incremental Training for SVM-Based Classification with Keyword Adjusting

This paper analyzed the theory of incremental learning of SVM (support vector machine) and pointed out it is a shortage that the support vector optimization is only considered in present research of SVM incremental learning. According to the significance of keyword in training, a new incremental training method considering keyword adjusting was proposed, which eliminates the difference between incremental learning and batch learning through the keyword adjusting. The experimental results show that the improved method outperforms the method without the keyword adjusting and achieve the same precision as the batch method. Key words SVM (support vector machine) - incremental training - classification - keyword adjusting CLC number TP 18 Foundation item: Supported by the National Information Industry Development Foundation of ChinaBiography: SUN Jin-wen (1972-), male, Post-Doctoral, research direction: artificial intelligence, data mining and system integration.

不平衡数据下基于SVM增量学习的指挥信息系统状态监控方法

针对指挥信息系统历史状态样本有限的特点,基于支持向量机(support vector machines,SVM)设计了一种面向不平衡数据的SVM增量学习方法。针对系统正常/异常状态样本不平衡的情况,首先利用支持向量生成一部分新样本,然后通过分带的思想逐带产生分布更加均匀的新样本以调节原样本集的不平衡比。针对系统监控实时性要求高且在运行过程中会有新样本不断加入的特点,采用增量学习的方式对分类模型进行持续更新,在放松KKT(Karush-Kuhn-Tucker)更新触发条件的基础上,通过定义样本重要度并引入保留率和遗忘率的方式减少了增量学习过程中所需训练的样本数量。为了验证算法的有效性和优越性,实验部分在真实系统中获得的数据集以及UCI数据集中3类6组不平衡数据集中与现有的算法进行了对比。结果表明,所提算法能够有效实现对不平衡数据的增量学习,从而满足指挥信息系统状态监控的需求。

Prediction of Protein Structural Classes Using the Theory of Increment of Diversity and Support Vector Machine

Based on the concept of the pseudo amino acid composition （PseAAC）, protein structural classes are predicted by using an approach of increment of diversity combined with support vector machine （ID-SVM）, in which the dipeptide amino acid composition of proteins is used as the source of diversity. Jackknife test shows that total prediction accuracy is 96.6% and higher than that given by other approaches. Besides, the specificity （Sp） and the Matthew＇s correlation coefficient （MCC） are also calculated for each protein structural class, the Sp is more than 88%, the MCC is higher than 92%, and the higher MCC and Sp imply that it is credible to use ID-SVM model predicting protein structural class. The results indicate that： 1 the choice of the source of diversity is reasonable, 2 the predictive performance of IDSVM is excellent, and3 the amino acid sequences of proteins contain information of protein structural classes.

火箭发动机故障检测的快速增量单分类支持向量机算法

为解决液体火箭发动机故障诊断正负样本不平均问题,以及实现发动机稳态工作段自适应故障检测,建立了基于快速增量单分类支持向量机的异常检测模型。采取特征工程方法,对传感器获得的多变量时间序列进行特征提取。通过增量学习方法,对单分类支持向量机模型进行改进,并应用于液体火箭发动机异常检测,使单分类支持向量机检测模型具备对不同台次、不同工况的自适应性,提高了模型的计算速度。对多台次热试车数据的分析结果表明,该模型十分有效,训练速度快,具备实用价值。

脑机接口中基于BISVM的EEG分类

针对脑电信号(EEG)分类问题,提出基于批处理增量式支持向量机(BISVM)的分类方法.将所有数据通过批处理进行分组,采用第1组数据在SVM中建立初始分类器模型,将剩余组内数据顺序作为新增样本,对满足卡罗需-库恩-塔克(KKT)条件的样本进行增量学习和减量去学习,不断判断KKT条件并更新参数,丢弃错误样本,对初始分类器模型进行更新.对2008年脑机接口竞赛数据及本实验室采集数据,用小波包分解(WPD)结合共空间模式(CSP)进行特征提取,SVM、ISVM及BISVM分类.结果表明,BISVM的平均分类准确率相对SVM及ISVM分别提高了3.3%及0.3%,BISVM平均训练时间相对ISVM从1.076s减少到0.793s.BISVM为改善计算机对大脑的适应性,实现快速实时在线的脑机接口系统奠定基础.

适于大规模数据集的块增量学习算法：BISVM

对支持向量机的大规模训练问题进行了深入研究,提出一种类似SMO的块增量算法。该算法利用increase和decrease两个过程依次对每个输入数据块进行学习,避免了传统支持向量机学习算法在大规模数据集情况下急剧增大的计算开销。理论分析表明新算法能够收敛到近似最优解。基于KDD数据集的实验结果表明,该算法能够获得接近线性的训练速率,且泛化性能和支持向量数目与LIBSVM方法的结果接近。

开集环境中基于增量学习的网络流量分类研究

面对网络流量新类别不断涌现的挑战,以及随之而来的开集识别和模型更新需求,文章提出了一种基于增量学习的开集网络流量分类方法。对于开集识别,支持向量机和K均值聚类算法的级联结构可以持续识别新类和已知类;对于模型更新,基于候选支持向量筛选的“样本回放”和新旧模型加权融合的“参数回放”方法,能有效解决“有类增量的灾难性遗忘”问题。与ISK和DACS方法相比,该方法应用在开集流量识别和分类任务中表现出显著优势,F1分数能提高1百分点至8百分点,分类速度也优于现有方法。

WSN中基于Mini Batch K-Means与SVM的入侵检测方案

无线传感器网络通常部署在复杂的户外环境,易遭受各种攻击。多数入侵检测系统均采用数据挖掘算法对网络数据包进行分析,但在处理大样本集时,其效率明显降低。针对这一缺点,提出一种基于Mini Batch K-Means和SVM的入侵检测方案。该方案首先分别对正常行为特征库和异常行为特征库进行Mini Batch K-Means聚类,取得类中心作为各类的代表样本并赋予权值,将其传入SVM分类器作为训练数据,得到分类超平面,通过该超平面对待测样本作出判断。解决了如K-Means、KNN、SVM等传统数据挖掘算法在大数据样本集数据分析中面临的低效问题。仿真结果表明,该方案能快速准确地判断样本类别,其检测率达到98.7%。与K-Means、KNN和SVM相比,不仅达到了同样高的检测率,而且明显提高了入侵检测的时间效率。

基于增量自适应支持向量机的AFM尖端磨损识别

为了提高纳米加工刀具磨损状态在线监测的精度与泛化能力,提出一种基于增量自适应支持向量机的基于原子力显微镜(AFM)尖端磨损识别方法。该方法以横向力的峰-峰值和方差作为特征变量,通过移动视窗获取增量数据;以维持Kuhn-Tucher定理所要求的最优化条件为准则,在当前支持向量机解结构基础上自适应修改正则化参数C和核参数σ,以获得更新支持向量机结构,并对增量数据及受其扰动的原数据进行分类;根据尖端失效点数量走势,判定尖端磨损程度。实验证明该算法在识别精度与时间上可满足在线检测要求。与定向非循环图支持向量分类器对比,该算法具有更强的鲁棒性与更高的泛化能力。

基于GA-SVM的多规格小批量生产的装配工时估算模型

对于制造承包商来说,在正式接收订单之前,为了指导报价和预测交货日期,有必要对工时(Man-hours,MH)进行评估。装配工时作为工时的重要组成部分,具有重要的实际研究意义。针对多规格、小批量生产的特点,提出了一种基于支持向量机(Support vector machine,SVM)的装配工时估算模型。除了单部件属性、装配过程和历史工时数据外,还考虑了可量化装配复杂性的最短路径长度平均值(Average of shortest path length,ASPL)作为装配MH的影响因素,并提出了基于Creo JLink三维模型的这些因素的自动计算方法。通过对几种算法的比较,选择SVM作为装配体MH建模的最优算法。将遗传算法(Genetic algorithm,GA)应用于SVM中,有利于在SVM中搜索最优参数c和g时避免了局部求解,加快了收敛速度。最后,对所提出的GA-SVM模型进行训练,并应用于雷达装置仿生腿的装配工时预测。实验结果表明,GA-SVM具有比本文其他方法更高的预测精度,整个预测过程仅需3 min左右。

基于增量式学习的电力监控系统网络安全攻击检测研究

为加强电力监控系统管控力度,确保电网平稳运行,以增量式学习技术为基础构建攻击检测模型。基于成批增量式学习算法,引入支持向量机,设计支持向量机下增量式学习算法;用二次规划问题界定支持向量机的分类超平面最优化目标函数;依据Karush Kuhn Tucker最优化条件,筛选增量集合,结合通用入侵检测模型,得到适用于电力监控系统网络的安全攻击检测模型。实验表明所建模型仅出现3次误检与1次漏检,检测耗时只有0.38 s,快速、精准地完成攻击检测,具备较高的效用性与可行性,能够保障网络安全。

基于机器学习的遥机械臂震颤消除

实现远程操作机器人和人类操作者的同步是一项有意义且具有挑战性的任务,其中需要克服的关键问题之一是消除人类手部的生理性震颤。为了解决这个问题,使用两种机器学习方法对人手震颤进行预测以及消除,并且对两种机器学习方法进行性能比较。对震颤消除器的数学模型进行分析并确定模型的输入及输出,建立了一个基于机器学习的消颤器。通过仿真实验,得到两种机器学习方法的性能评估指标,最终确定基于宽度-增量学习系统的消颤器更加有效且性能理想。仿真实验表明:基于宽度-增量学习系统的消颤器展现了很好的回归能力、泛化能力和快速学习能力。

基于HOG特征和TSVM算法的车标识别

支持向量机(Support Vector Machine,SVM)是建立车标识别模型的主要智能方法之一。考虑SVM存在计算复杂度高和无法实现增量学习等问题,提出一种基于孪生支持向量机(Twin SVM,TSVM)增量学习算法,并结合HOG特征设计一种车标识别系统。首先利用特征检测结合仿射变换技术,实现车标的精准定位;然后提取车标图像HOG特征,并通过对矩阵的逆运算进行分解和重组,实现TSVM增量学习。最后利用车标数据集训练分类模型,实现对车标的分类。实验结果表明,文中提出的算法在车标数据集上实现了91.77%的识别率,优于其他几种识别算法,证明了文中提出算法的有效性。