Topology distance for manifold clustering

Manifold clustering is a widely used techniques in pattern recognition and machine learning. It partition a set of input data into several clusters each of which contains data points from a separate, simple low-dimensional manifold. In order to cluster manifold, we propose a novel distance measure based on topology structure that can efficiently represent the underlying manifold. Under this distance measure, data points belong to the same clusters are more closed and that of the different clusters are farther apart. By using normalized cut on similarity matrix, clusters can be found with ease. Experiments on both synthetic data and real data show that our method is feasible and promising in manifold clustering.