Kernel semi-supervised graph embedding model for multimodal and mixmodal data
Kernel semi-supervised graph embedding model for multimodal and mixmodal data
复制标题
多模态和混合模态数据的核半监督图嵌入模型
DOI：
10.1007/s11432-018-9535-9
复制
发表时间：
2020
期刊：
影响因子：
--
通讯作者：
Chu Tianguang
中科院分区：
文献类型：
--
作者：
Zhang Qi;Li Rui;Chu Tianguang
亲爱的编辑，半监督学习在机器学习中得到了越来越多的关注，因为同时使用标记和非标记训练样本有助于提取区分特征，同时减少了耗时和劳动密集型的标记负担。为了在多模式(即同一类别的数据表现为单独的聚类)和混合模式(即来自不同类别的数据具有混合模式)数据[1]上提取特征，我们在[2]中提出了一种半监督图嵌入(SGE)模型，将软标签信息与数据的分层局部性结合起来。通过对类间加权可分性的极大化和类内局部距离的最小化处理，可以很好地捕捉具有多峰或混合峰分布特性的数据的内在特征，但由于SGE模型是一种线性技术，在刻画多峰和混合峰数据的非线性结构特征方面并不总是能得到令人满意的结果。根据核理论，当数据用核算子非线性映射到高维点积空间时，非线性降维问题可以有效地线性求解[3，4]。这激发了我们的兴趣--
Dear editor, Semi-supervised learning has obtained increasing interests in machine learning, because making use of both labeled and unlabeled training samples helps extracting discriminative features and meanwhile reduces the time-consuming and labor-intensive labeling burden. For extracting features upon multimodal (ie, data of the same class exhibits separate clustering) and mixmodal (ie, data from different classes has mixed modality) data [1], we have presented a semi-supervised graph embedding (SGE) model in [2] to incorporate the soft label information with hierarchical locality of data. Through the maximizing process upon the weighted between-class separability as well as the minimizing processes upon the localitypreserved within-class and scaled overall-class data distances respectively, the intrinsic characters of data with multimodal or mixmodal distributing properties can be well captured.However, as the SGE model is a linear technique, it might not always give satisfying results in capturing the nonlinear structural characteristics of multimodal and mixmodal data. According to the kernel theory, when data is mapped nonlinearly with a kernel operator into a highdimensional dot product space, the nonlinear dimensionality reduction problems can be efficiently solved linearly [3, 4]. This motivates us to ex-