摘要
图结构具有较强的表达能力和较高的灵活性,对图结构数据的识别及分类属于结构模式识别的范畴.对图结构数据的研究思路是将图结构数据通过图核转化为向量空间中的向量,然后采用传统的机器学习算法对其进行分析.基于图结构的数据表示与分析已经成为机器学习领域的研究热点.于是提出对经典的图结构分析方法进行扩充,利用核线性判别分析方法(KLDA)对图核诱导的高维特征空间进行降维,得到与原始图结构特征空间对应的低维度的特征空间,然后采用传统的机器学习算法对这些新的数据进行分析.在标准数据集上的实验效果验证了该方法的有效性.
Graph structure has strong expression ability and high flexibility.The identification and classification of graph structure data fall into the category of structural pattern recognition.The research idea of the graph structure data is to transform the graph structure data to the vector in the vector space,then the traditional machine learning algorithm is used to analyze the vector.Data representation and analysis based on graph structure has become a hot research topic in the field of machine learning.The classical graph kernel method was extended.The kernel linear discriminant analysis(KLDA)was employed to reduce the dimension of the high dimension feature space,and the low dimensional feature space corresponding to the original graph structure data was obtained.Then the traditional machine learning algorithm was used to analyze these new data.The effectiveness of the proposed method is verified by the experimental results on standard data sets.
基金
国家自然科学基金(61473149)资助
关键词
图分类
图核
核线性判别分析
降维
graph classification
graph kernel
kernel linear discriminant analysis
dimensionality reduction