An improved algorithm on spectral clustering

This paper raises an improved algorithm - spectral clustering algorithm based on particle swarm optimization, after a thorough study on the reasons of data input sensitivity of spectral clustering algorithm. The algorithm combines the Particle Swarm Optimization and spectral clustering algorithm. The overall convergence of this new algorithm not only overcomes the faults of locally optimal solutions and its sensitivity to the initial start centers in k-means algorithm, but also enjoys a higher converging speed.