基于重建系数的子空间聚类融合算法

Reconstruction coefficients based subspace clustering combination algorithm

针对稀疏子空间聚类(sparse subspace clustering,SSC)和低秩子空间聚类(low rank subspace clustering,LRSC)这两种子空间聚类方法的聚类准确率和稳定性不够高,提出一种基于重建系数的子空间聚类融合算法(reconstruction coefficients based subspace clustering combination algorithm,RCSCC)。该算法基于重建系数,将稀疏子空间聚类和低秩子空间聚类分别得到的相似度矩阵进行点乘融合运算,然后再用谱聚类来得到最后的聚类结果。实验结果表明,改进后的聚类融合算法不仅提高了聚类的准确率,还有效提高了聚类的稳定性和鲁棒性,从而验证了改进后的算法是有效可行的。

Aiming at the clustering accuracy and stability of SSC and LRSC are not high enough, this paper proposed a reconstruction coefficients based subspace clustering combination algorithm （ RCSCC）, which obtained the final similarity matrix based on point multiplication from the similarity matrixes got by sparse subspace clustering and low rank subspace clustering respectively, and then spectral clustering was done. Experimental results show that the improved algorithm can not only achieve higher accuracy, but also effectively improve the stability and robustness of clustering, which verifies that the improved algorithm is efficacious and feasible.