Cluster Trees on Manifolds
Balakrishnan, Sivaraman, Narayanan, Srivatsan, Rinaldo, Alessandro, Singh, Aarti, Wasserman, Larry
–Neural Information Processing Systems
We investigate the problem of estimating the cluster tree for a density $f$ supported on or near a smooth $d$-dimensional manifold $M$ isometrically embedded in $\mathbb{R}^D$. We study a $k$-nearest neighbor based algorithm recently proposed by Chaudhuri and Dasgupta. Under mild assumptions on $f$ and $M$, we obtain rates of convergence that depend on $d$ only but not on the ambient dimension $D$. We also provide a sample complexity lower bound for a natural class of clustering algorithms that use $D$-dimensional neighborhoods.
Neural Information Processing Systems
Dec-31-2013