Convex Set Disjointness, Distributed Learning of Halfspaces, and LP Feasibility

by   Mark Braverman, et al.

We study the Convex Set Disjointness (CSD) problem, where two players have input sets taken from an arbitrary fixed domain U⊆R^d of size | U| = n. Their mutual goal is to decide using minimum communication whether the convex hulls of their sets intersect (equivalently, whether their sets can be separated by a hyperplane). Different forms of this problem naturally arise in distributed learning and optimization: it is equivalent to Distributed Linear Program (LP) Feasibility– a basic task in distributed optimization, and it is tightly linked to Distributed Learning of Halfdpaces in R^d. In communication complexity theory, CSD can be viewed as a geometric interpolation between the classical problems of Set Disjointness (when d≥ n-1) and Greater-Than (when d=1). We establish a nearly tight bound of Θ̃(dlog n) on the communication complexity of learning halfspaces in R^d. For Convex Set Disjointness (and the equivalent task of distributed LP feasibility) we derive upper and lower bounds of Õ(d^2log n) and Ω(dlog n). These results improve upon several previous works in distributed learning and optimization. Unlike typical works in communication complexity, the main technical contribution of this work lies in the upper bounds. In particular, our protocols are based on a Container Lemma for Halfspaces and on two variants of Carathéodory's Theorem, which may be of independent interest. These geometric statements are used by our protocols to provide a compressed summary of the players' input.


page 13

page 27

page 29


On Communication Complexity of Classification Problems

This work introduces a model of distributed learning in the spirit of Ya...

The Minimax Complexity of Distributed Optimization

In this thesis, I study the minimax oracle complexity of distributed sto...

Fast Distributed Algorithms for LP-Type Problems of Bounded Dimension

In this paper we present various distributed algorithms for LP-type prob...

Property Testing of LP-Type Problems

Given query access to a set of constraints S, we wish to quickly check i...

Improved Communication Lower Bounds for Distributed Optimisation

Motivated by the interest in communication-efficient methods for distrib...

Efficient Protocols for Distributed Classification and Optimization

In distributed learning, the goal is to perform a learning task over dat...

Majorizing Measures for the Optimizer

The theory of majorizing measures, extensively developed by Fernique, Ta...