Learning Sparse Representations by Non-Negative Matrix Factorization and Sequential Cone Programming (Special Topic on Machine Learning and Optimization)
Matthias Heiler and Christoph Schnrr