Optimization and generalization guarantees for overparameterized low-rank matrix reconstruction

Concretely, we focus on the problem of reconstructing a low-rank matrix from a few measurements via a natural nonconvex formulation.