Reviews: Dropping Symmetry for Fast Symmetric Nonnegative Matrix Factorization

Summary: The authors solve an important NMF problem namely the symmetric NMF which arises in a wide variety of real world problems such as clustering in domains such as images, document analysis. They propose to rewrite the objective as a regular NMF problem with an additional regularization term of requiring the two matrix factors to be the same. This enables them to now apply standard NMF alternating update algorithms such as ANLS and HALS to solve the symmetric NMF problem. Real-world results are shown by experiments on CBCL, ORL, MNIST datasets which obtain qualitatively good results and also are pretty fast. Comments: The problem is well-defined and the approach/results are clearly presented.