arXiv · 1705.05348
Exploiting the Structure via Sketched Gradient Algorithms
Abstract
Sketched gradient algorithms have been recently introduced for efficiently solving the large-scale constrained Least-squares regressions. In this paper we provide novel convergence analysis for the basic method {\it Gradient Projection Classical Sketch} (GPCS) to reveal the fast linear convergence rate of GPCS towards a vicinity of the solution thanks to the intrinsic low-dimensional geometric structure of the solution prompted by constraint set. Similar to our analysis we observe computational and sketch size trade-offs in numerical experiments. Hence we justify that the combination of gradient methods and the sketching technique is a way of designing efficient algorithms which can actively exploit the low-dimensional structure to accelerate computation in large scale data regression and signal processing applications.
Explore related subjects
Keep this discovery
Junqi Tang, Mohammad Golbabaee, Mike Davies. 2017-05-15. Exploiting the Structure via Sketched Gradient Algorithms. https://arxiv.org/abs/1705.05348
Cite the original work for its findings. Save a collection to share your selection of sources.