NeurIPS 2020

Adaptive Reduced Rank Regression


Meta Review

The paper studies the low rank regression problem with a very small number of samples and a large number of features. For certain settings the paper gives improved results, and the algorithm is also arguably much simpler than existing algorithms. The authors also prove lower bounds showing optimality of their algorithms, and convincing experimental results. The algorithm is in fact so simple and for an important problem, that the reviewers were worried they might be missing some existing work, though we did not find any. Overall it's nice that the authors are able to say something fundamental about PCA.