A New Continuous Optimization Model for Spectral Clustering

Clustering and graph partition play an important role in big data analysis. These problems are often formulated as combinatorial optimization models. Consequently, to solve them efficiently is difficult. In this paper, we propose a novel continuous optimization model, and a block coordinate decent method to solve it. Numerical experiments show that the new approach has great potential in dealing with clustering and graph partition problems. We also give preliminary analysis on the relationship between our model and the original combinatorial optimization model.