A fast spectral cluster algorithm based on bipartite graph

Recently, spectral clustering has been applied successfully in a variety of different areas. While choosing the similarity graph and its parameters for spectral clustering is not a trivial task and its computational complexity is rather high. In this paper we proposed a fast bipartite graph spectral algorithm which uses evidence accumulation method and algebraic transformation method to avoid the above shortcoming. Experiments on real-world document sets demonstrate the effectiveness of the proposed algorithm.