您现在的位置: 首页> 研究主题> 多类支持向量机

多类支持向量机

多类支持向量机的相关文献在2002年到2022年内共计98篇,主要集中在自动化技术、计算机技术、无线电电子学、电信技术、金属学与金属工艺 等领域,其中期刊论文69篇、会议论文1篇、专利文献2601212篇;相关期刊51种,包括聊城大学学报(自然科学版)、系统工程与电子技术、数据采集与处理等; 相关会议1种,包括2010年全国模式识别学术会议(CCPR2010)等;多类支持向量机的相关文献由283位作者贡献,包括刘佳、吴石、张莉等。

多类支持向量机—发文量

期刊论文>

论文:69 占比:0.00%

会议论文>

论文:1 占比:0.00%

专利文献>

论文:2601212 占比:100.00%

总计:2601282篇

多类支持向量机—发文趋势图

多类支持向量机

-研究学者

  • 刘佳
  • 吴石
  • 张莉
  • 徐图
  • 李凡长
  • 杜培军
  • 沈兰荪
  • 王邦军
  • 代劲
  • 何大可

多类支持向量机

-相关会议

  • 期刊论文
  • 会议论文
  • 专利文献

搜索

排序:

年份

    • 张静; 刘忠宝; 宋文爱; 富丽贞; 章永来
    • 摘要: Support vector machine (SVM ) ,a typical classification method ,has been widely used in stellar spectra classification . It performs well in practice ,while it encounters the multi-class classification challenge .In order to solve the problem above , multi-class support vector machine (MCSVM) was proposed in this paper based on the in-depth analysis of SVM .Meanwhile , the stellar spectra classification model based on multi-class support vector machine was constructed .The advantage of the pro-posed method is that the samples'class can be determined by a classification process .Comparative experiments with the existed multi-class classification method on the SDSS DR8 datasets verify the effectiveness of the proposed method .%支持向量机作为一种经典的分类方法被广泛应用于恒星光谱分类领域.该方法在实际应用中取得了较为理想的分类效果,但其面临无法解决多分类问题的挑战.在支持向量机的基础上,提出多类支持向量机,建立基于多类支持向量机的恒星光谱分类模型.该方法的最大优势是经过一次分类过程,可以确定多类样本的类属.SDSS DR8恒星光谱数据上的比较实验表明,本研究所提的方法较之已有多分类方法在分类性能上有一定的提升.
    • 扈常生; 黄永烈; 顾斯洋; 崔鹏; 张英; 王军
    • 摘要: 作为电力系统中电能转换和传输的核心设备,保证变压器的安全稳定运行至关重要.变压器发生故障会造成电能传输中断,影响整个电气系统的稳定运行,造成巨大的经济损失.基于多维特征量的电力变压器诊断技术的应用可以精准检测变压器异常故障,保证变压器稳定运行.因此,分析智能算法变压器的故障诊断方法,并结合多维特征量模型的特点,研究基于多维特征量的多层次信息融合故障诊断技术.
    • 姚湘; 徐平平; 王华君
    • 摘要: 针对一般基于知识迁移的方法对未知视角不可用和难以扩展新数据的问题,提出一种基于非线性模型的无监督学习方法,即基于非线性知识迁移(nonlinear knowledge shift,NKS)的串联特征学习.提取密集动作轨迹,并利用通用码书编码;提取动作捕捉数据模拟点的密集轨迹,产生一个仿真数据的大型语料库来学习NKS,其中,轨迹提取前在视角方向上投影模拟点;再从真实视频中提取轨迹,用于训练和测试表示学习过程的轨迹,利用多类支持向量机分类串联特征.在两大通用人体动作识别数据库IXMAS和3D(N-UCLA)上验证了该方法的有效性,实验结果表明,在IXMAS数据集、不同摄像机情况下,该方法的识别精度高于同类方法至少3.5%,在3D(N-NCLA)数据集、双摄像头情况下,识别精度至少提高4.4%.在大部分动作识别中也取得最佳识别率,此外,该方法的训练时间可忽略不计,有望应用于在线人体动作识别系统.
    • 王剑; 张伟华; 李跃新
    • 摘要: 为提高多类支持向量机分类器对多目标的分类准确度,提出一种结合无向图模型优化的多类支持向量机分类器.首先,利用余弦测度计算训练数据之间的相似度,构建包含训练数据和相似度矩阵的无向图模型,求解相似度约束矩阵.然后,将相似度约束矩阵引入多类支持向量机求解的目标函数,构建优化的多类支持向量机分类器.最后,将优化的多类支持向量机分类器用于智能交通领域,结合梯度方向直方图特征检测行人和车辆目标.实验表明,该方法检测行人和车辆目标的错误率低于经典的多类支持向量机分类器和目前主流的目标检测方法.
    • 何俊林; 赵晓亮; 孙连海; 甘胜江
    • 摘要: 提出一种人体行为识别方法.构建MACH滤波器组,对视频片段的三维时空体进行滤波,得到时空相关体;采用三层最大池化方法提取时空相关体的特征向量,采用高斯隶属函数对池化特征向量进行扩展;构建多类SVM分类器并进行特征分类,识别行为类别.在ADL和UCF Sports两个国际上通用的人体行为数据集上进行人体行为识别实验,实验结果表明,该方法的识别率高于现有的人体行为识别方法,对不同人体行为的区分能力更强.%A human activity recognition method was proposed.A set of MACH filters was built to filter 3D space-time volume of video clips,for obtaining a space-time correlation volume.Three-layer max-pooling method was used to extract feature vector of the space-time correlation volume,and Gauss membership function was used to extend the pooling feature vector.Multi-class SVM classifier was built to classify features and recognize activities.Experimental results on two internationally common datasets of human activity including ADL and UCF Sports show that,the proposed method has higher recognition rate than the existing methods for human activity recognition,and it has strong ability to distinguish different human activities.
    • 李远成; 刘斌
    • 摘要: 针对传统网络流量分类方法准确率不高、开销较大且应用领域受限等诸多问题,文中提出一种基于主动学习支持向量机的网络流量分类方法.该方法采用基于OVA方法的多类支持向量机来进行分类,首先,针对支持向量机参数选择,提出了一种改进的网格搜索法来寻求最优参数;然后,为了降低需要标注的样本数,提出一个改进的启发式主动学习样本查询准则;最后,基于上述方法构造基于主动学习的多类支持向量机分类器.结果表明,该方法可以在需要标注的样本数非常少的情况下明显提高网络流量分类的准确率和效率,仅需传统方法所需11%的样本数即可达到98.7%的分类准确率.%Aiming to solve the problems such as low accuracy,large overhead and limitation of applications for traditional network traffic classification,this paper presents a novel network traffic classification method based on active learning support vector machine(ALSVM).This method applied the OVA method to build multi-class support vector machine.Firstly,we propose an improved grid search method to seek the optimal parameter for ALSVM.Then,an improved heuristic active learning sample query criteria is proposed to reduce the number of label samples.Lastly,an active multi-class support vector machine classifier is constructed for network traffic classification.Experimental results show that,this method can significantly improve the accuracy and efficiency of network traffic classification with much fewer label samples.We can gain 98.7% classification accuracy with 11% of the conventional method required number of label samples.
    • 刘晟; 彭宗举; 陈嘉丽; 陈芬; 郁梅; 蒋刚毅
    • 摘要: 深度视频编码中最优深度划分和模式选择过程具有非常高的计算复杂度。提出了基于多类支持向量机(MSVM, multi-class support vector machine)的深度视频帧内编码快速算法。该算法包括离线模型训练和快速编码2个部分。在离线模型训练中,用深度视频最大编码单元(LCU, largest coding unit)的最优划分深度作为标签,当前LCU的空域复杂度、空域相邻LCU的最优划分深度和彩色视频对应LCU的最优划分深度作为特征去构造MSVM模型。在编码时,提取LCU的特征,根据MSVM模型得到划分深度的预测值。根据该预测值提前终止编码单元递归划分和模式选择过程。实验结果表明,提出的算法在几乎不影响虚拟视点质量的情况下,平均节省35.91%的总体编码时间和40.04%的深度编码时间。%The recursive splitting process of largest coding unit (LCU) and the mode search process of coding unit imposed enormous computational complexity on encoder. A multi-class support vector machine-based (MSVM) fast coding unit (CU) size decision algorithm for 3D-HEVC depth video intra-coding was proposed. The algorithm included two steps:off-line training and fast CU size and mode decision. In the process of off-line training, a MSVM model was constructed, where the texture complexity of current LCU, the optimal partition depth of its spatial neighboring LCU and co-located LCU in texture video were treated as feature vectors, and the optimal partition depth of LCU was utilized as corresponding class label. In the process of fast CU size and mode decision, features of LCU were extracted before cod-ing a LCU, then, a MSVM model was used to predict the class label. Finally, the class label that represents the largest parti-tion depth of the current LCU was employed to terminate the CU recursive splitting process and CU mode search process. Experimental results show that the proposed algorithm saves the encoding time of 3D-HEVC by 35.91%on average, and the encoding time of depth video by 40.04%on average, with negligible rendered virtual view image degradation.
    • 练瑶; 黄泽炽; 葛猛
    • 摘要: Objective:B-cell epitopes (BCEs)can be divided into subclasses according to the isotype of antibody (Ab)that BCEs can induce.It is important to discriminate among these BCEs subclasses because it would be beneficial to understand why the immune system produces different Ab isotypes against different BCEs.Based on a multi-class sup-port vector machine (SVM),models have been developed to discriminate among BCEs subclasses.Methods:A four-class dataset including four BCEs subclasses was created to train and test the models.Various primary sequence features including amino acid composition,quasi-sequence-order and dipeptide composition have been computed for comparing their ability to discriminate BCEs subclasses using five-fold cross validation.Results:It was observed that dipeptide composition based model achieved the highest overall accuracy of 61 .58%.Moreover,a web server,BCESCP,of the best performing multi-class classification model has been provided for predicting subclass type of BCEs.BCESCP is free-ly available at http://www.bioinfo.tsinghua.edu.cn /epitope /BCESCP/.%目的:根据诱导的特异性抗体种型,B 细胞表位被分成不同的亚类。探索表位多类亚类之间的区别非常重要,能促进揭示免疫系统为什么会针对不同的表位产生特异性抗体应答。基于多类支持向量机,发展一个能区分多类表位亚类并且能预测 B 细胞表位的亚类类别的模型。方法:训练模型的数据集来源于免疫表位数据库,数据集包含4类数据,对应4种 B 细胞表位亚类:IgA 表位,IgE 表位,IgG 表位以及 IgM表位。通过5折交叉验证,分别探索氨基酸组成特征,quasi-序列顺序特征以及二肽组成特征区分表位多类亚类的能力。结果:实验结果表明二肽组成特征的区分性能最好,整体准确率为61.58%,应用此多类分类模型,开发了一个名为BCESCP 的免费使用的 B 细胞表位的亚类类别预测服务器,BCESCP 可以通过如下地址访问:http://www.bioin-fo.tsinghua.edu.cn /epitope /BCESCP/。
    • 赵振华; 姜大立; 刘军; 薛皓文
    • 摘要: 随着军民融合式后勤保障体制改革的深入推进,多种储备模式并存是战备物资储备发展的大趋势.针对具体物资,选择何种储备模式是需要解决的首要问题.针对该问题,建立了影响物资储备模式选择的特性指标集,通过问卷调查取得特性指标向量和推荐储备模式,利用模糊聚类方法消除由于主观因素对储备模式选择带来的影响,然后将聚类结果用于训练与测试多类支持向量机,从而形成一种实现储备模式自动选择的机器.通过数据验证,取得了较好的实验结果,实现了具体战备物资储备模式的有效选择.
    • 刘斌; 张晓婧; 徐谦
    • 摘要: 为了解决数据分类效率问题,引入了并行概念学习的方法,针对二叉树分类法的“累积误差”问题,提出了一种并行概念的描述语言.在此基础上,设计了基于二叉树的多类SVM并行分类系统.通过对系统性能的分析表明,在样本空间庞大、样本涉及属性广的数据分类时,该系统具有很好的分类效果.
  • 查看更多

客服邮箱:kefu@zhangqiaokeyan.com

京公网安备:11010802029741号 ICP备案号:京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有
  • 客服微信

  • 服务号