Regularizers versus Losses for Nonlinear Dimensionality Reduction: A Factored View with New Convex Relaxations

Open in new window