Manifold Learning via Manifold Deflation

Daniel Ting, Michael I. Jordan

Nonlinear dimensionality reduction methods provide a valuable means to visualize and interpret high-dimensional data. However, many popular methods can fail dramatically, even on s…