Links
Slides
Modern datasets usually have a large number of features or training samples, and they are usually stored in a distributed manner. Motivated by the need of solving sparse optimization problems with large datasets, we propose two approaches including (i) distributed implementations of prox-linear algorithms and (ii) GRock, a parallel greedy coordinate descent method.
P-FISTA: parallel fast iterative soft-thresholding algorithm
PD-ADMM: parallel dual alternating direction method of multipliers
Z. Peng, M. Yan, and W. Yin. Parallel and Distributed Sparse Optimization, preprint, 2013