The objective in Liu Ce's project is to learn a Gibbs model of face in a high dimensional space, like texture modeling, by learning more and more (K) feature statistics from a training set, the face model starts to render random faces that resemble faces in the training set. As K increases, the faces become more realistic.

C. Liu, S. C. Zhu, and H. Y. Shum, "Learning inhomogeneous Gibbs models of faces by minimax entropy," Proc Int'l Conference on Computer Vision, Vancouver, Canada, July, 2001.