We propose a novel fr a mework for thedeterministicconstruction oflinear, near-isometric embeddings of a nite set of data points. -We demonstrate that our fr a mework is useful for a number of applications in machine learning and signal processi