Visualization of Oja's rule application as a method to find principal components of a set of example pictures of faces. Then computed components, so called eigenfaces, are used as a basis for subspace on which few projections are computed. Look at what happens when pictures not used at the begining, during PC computation, are projected on this subspace.
And yes, it's me on these pictures. I prepared these examples around 5 years ago. Video was written in C with SDL and OpenGL.