Authors: R. Cabral (Carnegie Mellon University,USA) , F. De la Torre (Carnegie Mellon University,USA), J. Costeira (Instituto Superior Técnico, Portugal) A. Bernardino (Instituto Superior Técnico, Portugal)
Chapter Description
In this chapter, we study the topic of complexity penalization for visual learning tasks through rank minimization models. The use of rank criteria has been pervasive in computer vision applications as a mean of exploiting physical constraints of a model or to minimize its complexity, be it in degrees of freedom or in data redundany. All these problems are directly or indirectly related to the problem of recovering a rank-k matrix Z data from a corrupted data matrix X.