Material Detail

This video was recorded at NIPS Workshops, Whistler 2010. Online prediction methods are typically presented as serial algorithms running on a single processor. However, in the age of web-scale prediction problems, it is increasingly common to encounter situations where a single processor cannot keep up with the high rate at which inputs arrive. In this work we present the distributed mini-batch algorithm, a method of converting any serial gradient-based online pre- diction algorithm into a distributed algorithm that scales nicely to multiple cores, clusters, and grids. We prove a regret bound for this method that is asymptotically optimal for smooth convex loss functions and stochastic inputs. Moreover, our regret analysis explicitly takes into account communication latencies that occur on the network. Our method can also be adapted to optimally solve the closely- related distributed stochastic optimization problem.

Disciplines:

• Science and Technology  / Computer Science  / Programming & Programming Languages