Spectral Regularization Algorithms for Learning Large Incomplete Matrices

We use convex relaxation techniques to provide a sequence of regularized low-rank solutions for large-scale matrix completion problems. Using the nuclear norm as a regularizer, we provide a simple ...