On the relation between the randomized extended Kaczmarz algorithm and coordinate descent

Abstract

In this note we compare the randomized extended Kaczmarz (EK) algorithm and randomized coordinate descent (CD) for solving the full-rank overdetermined linear least-squares problem and prove that CD needs fewer operations for satisfying the same residual-related termination criteria. For the general least-squares problems, we show that first running CD to compute the residual and then standard Kaczmarz on the resulting consistent system is more efficient than EK.