The \emph{Undecided-State Dynamics} is a well-known protocol for distributed consensus. We analyze it in the parallel \pull\ communication model on the complete graph for the \emph{binary} case (every node can either support one of \emph{two} possible colors, or be in the undecided state). An interesting open question is whether this dynamics \emph{always} (i.e., starting from an arbitrary initial configuration) reaches consensus \emph{quickly} (i.e., within a polylogarithmic number of rounds) in a complete graph with n nodes. Previous work in this setting only considers initial color configurations with no undecided nodes and a large \emph{bias} (i.e., Θ(n)) towards the majority color. In this paper we present an \textit{unconditional} analysis of the Undecided-State Dynamics that answers to the above question in the affirmative. We prove that, starting from \textit{any} initial configuration, the process reaches a monochromatic configuration within O(logn) rounds, with high probability. This bound turns out to be tight. Our analysis also shows that, if the initial configuration has bias Ω(nlogn), then the dynamics converges toward the initial majority color, with high probability.
@article{arxiv.1707.05135,
title = {A Tight Analysis of the Parallel Undecided-State Dynamics with Two Colors},
author = {Andrea E. F. Clementi and Luciano Gualà and Francesco Pasquale and Giacomo Scornavacca and Emanuele Natale and Mohsen Ghaffari},
journal= {arXiv preprint arXiv:1707.05135},
year = {2018}
}