Compressed Sensing: RandomFaces prove that faces lie in a low dimensional manifold

It was only a matter of time, the Machine Learning community is finding ways to use compressed sensing following the work of Richard Baraniuk and Michael Wakin on Random projections of smooth manifolds.
There were a laplacianfaces, eigenfaces, we now have randomfaces, the projection of faces onto random bases. The work of Yi Ma ([1] [2] [3])seems to show that one can get similar results in terms of classification as those obtain from the more complex techniques producing the Laplacianfaces and other Eigenfaces. Some people would say that it may parallel what is happening in the primary visual cortex.

[ Des presentations sur le compressed sensing (acquisition comprimee) se feront a l'Institut Henri Poincaré (Paris, France) en Mars 2008, Méthodes variationnelles et parcimonieuses en traitement des signaux et des images , Merci Laurent]

References:
[1] Robust Face Recognition via Sparse Representation, John Wright, Arvind Ganesh, Allen Yang and Yi Ma. Submitted to IEEE Transactions on Pattern Analysis and Machine Intelligence. [Updated January 2008]
[2] Feature Selection in Face Recognition: A Sparse Representation Perspective, Allen Yang, John Wright, Yi Ma and Shankar Sastry. Submitted to IEEE Transactions on Pattern Analysis and Machine Intelligence.
[3] Computation and Relaxation of Conditions for Equivalence between L1 and L0 Minimization, Yoav Sharon, John Wright and Yi Ma. Submitted to IEEE Transactions on Information Theory.