Timezone: »

Statistically Preconditioned Accelerated Gradient Method for Distributed Optimization
Hadrien Hendrikx · Lin Xiao · Sebastien Bubeck · Francis Bach · Laurent Massoulié

Wed Jul 15 11:00 AM -- 11:45 AM & Wed Jul 15 10:00 PM -- 10:45 PM (PDT) @ None #None

We consider the setting of distributed empirical risk minimization where multiple machines compute the gradients in parallel and a centralized server updates the model parameters. In order to reduce the number of communications required to reach a given accuracy, we propose a preconditioned accelerated gradient method where the preconditioning is done by solving a local optimization problem over a subsampled dataset at the server. The convergence rate of the method depends on the square root of the relative condition number between the global and local loss functions. We estimate the relative condition number for linear prediction models by studying uniform concentration of the Hessians over a bounded domain, which allows us to derive improved convergence rates for existing preconditioned gradient methods and our accelerated method. Experiments on real-world datasets illustrate the benefits of acceleration in the ill-conditioned regime.

Author Information

Hadrien Hendrikx (INRIA)
Lin Xiao (Microsoft Research)
Sebastien Bubeck (Microsoft Research)
Francis Bach (INRIA - Ecole Normale Supérieure)
Laurent Massoulié (MSR-INRIA Joint Center)

More from the Same Authors