Matrix Recipes for Hard Thresholding Methods

摘要: In this paper, we present and analyze a new set of low-rank recovery algorithms for linear inverse problems within the class hard thresholding methods. We provide strategies on how to up these via basic ingredients different configurations achieve complexity vs. accuracy tradeoffs. Moreover, study acceleration schemes memory-based techniques randomized, ∈-approximate matrix projections decrease computational costs in process. For most configurations, theoretical analysis that guarantees convergence under mild problem conditions. Simulation results demonstrate notable performance improvements as compared state-of-the-art both terms reconstruction complexity.