Core Concepts
The convergence time of asynchronous opinion dynamics in social networks is bounded by O(n|E|^2(ε/δ)^2), where n is the number of agents, |E| is the number of edges in the social network, ε is the confidence bound, and δ is the stability parameter.
Abstract
The paper studies the convergence time of asynchronous opinion dynamics in social networks, where agents update their opinions one at a time in a random order. The key findings are:
The authors prove that the opinion dynamics are guaranteed to converge to a δ-stable state, where the length of each edge in the influence network is at most δ.
For arbitrary social networks, they provide an upper bound of O(n|E|^2(ε/δ)^2) on the expected convergence time, where n is the number of agents, |E| is the number of edges in the social network, ε is the confidence bound, and δ is the stability parameter.
For the special case of complete social networks, the authors show an even tighter upper bound of O(n^3(n^2 + (ε/δ)^2)), which significantly improves the previously best known bound of O(n^9(ε/δ)^2).
The authors also provide matching lower bounds, demonstrating the tightness of their analysis. Specifically, they construct a family of instances where the expected potential drop per step is exactly of the same order as the upper bound.
The results show that the convergence time in the asynchronous setting is of the same order as the best known bounds for the synchronous setting, despite the additional challenges posed by the asynchronous updates.
Stats
The paper does not contain any explicit numerical data or statistics. The key results are the upper and lower bounds on the convergence time of the asynchronous opinion dynamics.
Quotes
"The expected convergence time to a δ-stable state under uniform random asynchronous updates is O(Φ(S0)n|E|/δ^2) ≤ O(n|E|^2 (ε/δ)^2)."
"For the complete social network, the expected convergence time to a δ-stable state is at most O(n^3 (n^2 + (ε/δ)^2))."