Randomized Sparse Block Kaczmarz as Randomized Dual Block-Coordinate Descent
Apr 22, 2017
About this article
Published Online: Apr 22, 2017
Page range: 129 - 149
Received: Jan 01, 2015
Accepted: Mar 01, 2015
DOI: https://doi.org/10.1515/auom-2015-0052
Keywords
© 2017
This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License.
We show that the Sparse Kaczmarz method is a particular instance of the coordinate gradient method applied to an unconstrained dual problem corresponding to a regularized ℓ1-minimization problem subject to linear constraints. Based on this observation and recent theoretical work concerning the convergence analysis and corresponding convergence rates for the randomized block coordinate gradient descent method, we derive block versions and consider randomized ordering of blocks of equations. Convergence in expectation is thus obtained as a byproduct. By smoothing the ℓ1-objective we obtain a strongly convex dual which opens the way to various acceleration schemes.