Sparse Subspace Clustering for Stream Data

In the past few years, sparse subspace clustering (SSC) has gained many studies and found wide applications. However, SSC suffers from limitation in scalability. Furthermore, current methods could hardly tackle stream data where structure of subspaces may change along time. this paper, we propose a novel method to extend (StreamSSC). Our is based on maintaining small subset representatives characterize underlying during data. StreamSSC efficient both computation memory. Experimental results synthetic real-world streams demonstrate effectiveness StreamSSC. For efficiency, faster than existing online by roughly magnitude.