Scientists working with large volumes of high-dimensional data, such as global climate patterns, stellar spectra, or human gene distributions, regularly con- front the problem of dimensionality reduction: Þnding meaningful low-dimen- sional structur