作者: Jun Jiao , Xuan Mo , Chen Shen
DOI: 10.1007/978-3-642-11301-7_82
关键词:
摘要: In recent years, clustering techniques have become a useful tool in exploring data structures and been employed broad range of applications. this paper we derive novel image approach based on sparse representation model, which assumes that each instance can be reconstructed by the linear combination other instances. Our method characterizes graph adjacency structure weights coefficients computed solving l1-minimization. Spectral algorithm using these as weight matrix is then used to discover cluster structure. Experiments confirmed effectiveness our approach.