我们与稀少的支持向量机器(SVM ) 在特征选择学习策略。最近,所谓的 L <sub> p </sub>-SVM (0 p L <sub>1</sub>-SVM。然而, L <sub> p </sub>-SVM 是一非凸并且 non-Lipschitz 优化问题。数字...我们与稀少的支持向量机器(SVM ) 在特征选择学习策略。最近,所谓的 L <sub> p </sub>-SVM (0 p L <sub>1</sub>-SVM。然而, L <sub> p </sub>-SVM 是一非凸并且 non-Lipschitz 优化问题。数字地解决这个问题是挑战性的。在这份报纸,我们再用形式表示 L <sub> p </sub>-SVM 进有线性客观功能和光滑的限制(LOSC-SVM ) 的一个优化模型以便它能被数字方法为光滑的抑制优化解决。我们人工的数据集的数字实验显示出那 LOSC-SVM (0 p L <sub>1</sub>-SVM。展开更多
Protein phosphorylation/dephosphorylation is the central mechanism of post-translational modification which regulates cellular responses and phenotypes. Due to the efficiency and resource constraints of the in vivo me...Protein phosphorylation/dephosphorylation is the central mechanism of post-translational modification which regulates cellular responses and phenotypes. Due to the efficiency and resource constraints of the in vivo methods for identifying phosphorylation sites, there is a strong motivation to computationally predict potential phosphorylation sites. In this work, we propose to use a unique set of features to represent the peptides surrounding the amino acid sites of interest and use feature selection support vector machine to predict whether the serine/threonine sites are potentially phosphorylable, as well as selecting important features that may lead to phosphorylation. Experimental results indicate that the new features and the prediction method can more effectively predict protein phosphorylation sites than the existing state of the art methods. The features selected by our prediction model provide biological insights to the in vivo phosphorylation.展开更多
将语种和说话人识别的方法应用到英语发音错误检测系统,提出一种基于广义线性区分序列支持向量机(Generalized linear dis-criminant sequence based SVM,GLDS-SVM)的发音错误检测方法.主要创新点为:1)提出一种基于状态拼接的特征规整方...将语种和说话人识别的方法应用到英语发音错误检测系统,提出一种基于广义线性区分序列支持向量机(Generalized linear dis-criminant sequence based SVM,GLDS-SVM)的发音错误检测方法.主要创新点为:1)提出一种基于状态拼接的特征规整方案,增强SVM对发音特征的建模能力;2)提出一种基于多模型融合的模型训练策略,该策略可以更加充分地利用训练数据,并在一定程度上解决了由于真实发音错误数据缺乏造成的正负样本不均衡的问题;3)将GLDS-SVM与基于通用背景模型GMM(Universal background modelsbased GMM,GMM-UBM)的方法进行融合,以进一步提高发音检错性能.GLDS-SVM和GMM-UBM的融合系统在仿真测试集和真实测试集上的等错误率(Equal error rate,EER)分别达到9.92%和16.35%.同时,GLDS-SVM在模型占用空间和运算速度方面均比传统径向基函数(Radial basic function,RBF)核方法具有明显优势.展开更多
基金This work is supported in part by the National Natural Science Foundation of China under Grant Nos. 61502159, 61379057, 11101081, and 11271069, and the Research Foundation of Central South University of China under Grant No. 2014JSJJ019.
文摘我们与稀少的支持向量机器(SVM ) 在特征选择学习策略。最近,所谓的 L <sub> p </sub>-SVM (0 p L <sub>1</sub>-SVM。然而, L <sub> p </sub>-SVM 是一非凸并且 non-Lipschitz 优化问题。数字地解决这个问题是挑战性的。在这份报纸,我们再用形式表示 L <sub> p </sub>-SVM 进有线性客观功能和光滑的限制(LOSC-SVM ) 的一个优化模型以便它能被数字方法为光滑的抑制优化解决。我们人工的数据集的数字实验显示出那 LOSC-SVM (0 p L <sub>1</sub>-SVM。
文摘Protein phosphorylation/dephosphorylation is the central mechanism of post-translational modification which regulates cellular responses and phenotypes. Due to the efficiency and resource constraints of the in vivo methods for identifying phosphorylation sites, there is a strong motivation to computationally predict potential phosphorylation sites. In this work, we propose to use a unique set of features to represent the peptides surrounding the amino acid sites of interest and use feature selection support vector machine to predict whether the serine/threonine sites are potentially phosphorylable, as well as selecting important features that may lead to phosphorylation. Experimental results indicate that the new features and the prediction method can more effectively predict protein phosphorylation sites than the existing state of the art methods. The features selected by our prediction model provide biological insights to the in vivo phosphorylation.