Dynamics of Randomized Row-Action Methods for High-Dimensional Estimation
De Yue Lu
In this talk, I will present an exact analysis of the dynamics of randomized iterative methods for solving inference problems. I will show that, in the large systems limit, the dynamics of these algorithms converges to trajectories governed by a set of deterministic and coupled ODEs or PDEs. Analyzing these deterministic ODEs and PDEs allows one to establish performance guarantees of the associated randomized iterative algorithms.