斜投影核鉴别器的增量学习:理论及算法
首发时间:2012-10-29
摘要:分类器设计在模式识别中具有重要的地位,可以归结为函数逼近问题。先前我们在定义于模式特征空间上的再生核希尔伯特空间中,应用正投影准则研究了用于模式特征表示和鉴别的核非线性分类器设计问题。进一步,我们在此利用斜投影将某模式类别从其它类别中鉴别开来,得到斜投影核鉴别器,并设计对应的增量学习算法,以解决分类器在线训练问题和参数稀疏化问题。
关键词: 模式识别 核非线性分类器 核鉴别器 斜投影 增量学习
For information in English, please click here
Incremental learning of oblique projection-based kernel discriminator: Theory and algorithm
Abstract:Classifier design plays a paramount role in pattern recognition. Previously, we set the problem in theframework of function approximation, wherein a classifier is assumed to be an element of a Reproducing Kernel HilbertSpace (RKHS) continuously defined on the pattern feature space, and adopted orthogonal projection criteria for patternfeature representation and discrimination. In this manuscript, we adopt oblique projection to discriminate a pattern class, called the target class,from other classes, by obliquely projecting a pattern feature vector onto the subspace spanned by the training patternfeatures of the target class, along the subspace spanned by those of other classes, so that a type of nonlinearclassifier called Kernel Discriminator via Oblique Projection (KDOP) is obtained. In addition, we provide an incrementallearning algorithm for online training and sparsification of the KDOP.
Keywords: Pattern recognition kernel-based nonlinear classifier kernel discriminator oblique projection incremental learning
论文图表:
引用
No.****
同行评议
共计0人参与
勘误表
斜投影核鉴别器的增量学习:理论及算法
评论
全部评论0/1000