Robust Algorithms for Low-Rank and Sparse Matrix Models