A customizable Swiss Roll data set, with pictures and Matlab code. The original data comes from a four-component Gaussian Mixture Model.
A speech frames data set from TIMIT (DR1). Created for a phoneme acquisition experiment.