Scalable Structured Low Rank Matrix Optimization Problems

We consider a class of structured low rank matrix optimization problems. We represent the desired structure by a linear map, termed mutation, that can encode matrices having entries partitioned into known disjoined groups. Our interest arises in particular from concatenated block-Hankel matrices that appear in formulations for input-output linear system identification problems with noisy and/or partially unobserved data. We present an algorithm and test it against an existing alternative.