Tuesday, December 23, 2025

The Communication Complexity of Distributed Estimation


We examine an extension of the usual two-party communication mannequin by which Alice and Bob maintain chance distributions pp and qq over domains XX and YY, respectively. Their aim is to estimate

Exp,yq[f(x,y)]mathbb{E}_{x sim p, y sim q}[f(x, y)]

to inside additive error εvarepsilon for a bounded perform ff, recognized to each events. We check with this because the distributed estimation downside. Particular circumstances of this downside come up in a wide range of areas together with sketching, databases and studying. Our aim is to know how the required communication scales with the communication complexity of ff and the error parameter εvarepsilon.

The random sampling strategy — estimating the imply by averaging over O(1/ε2)O(1/varepsilon^2)

Related Articles

Latest Articles