Researchers crack voting speed limits in large-scale distributed networks
New paper proves consensus can form 10x faster in expander graphs…
Deep Dive
An analysis of the discrete incremental voting process (DIV) on undirected graphs. If the graph has conductance Φ, ratio of average to smallest degree γ, and maximal initial opinion difference K, the expected convergence time is O(n(K log(Kn) + γ n)/Φ²). This bound is essentially optimal for a large class of graphs of bounded expansion. For regular graphs, if the second largest eigenvalue is o(1/log² n) and K is o(n/log² n), then with high probability DIV converges to the initial average opinion (rounded up or down).
Key Points
- DIV process converges in O(n(K log(Kn) + γ n)/Φ²) steps on general graphs — optimal for bounded expansion.
- For regular expanders (second eigenvalue < o(1/log² n)), DIV reaches the initial average opinion w.h.p.
- Improves prior O(PODIS '23) bounds and gives first tight analysis for expander graphs.
Why It Matters
Tighter bounds for distributed voting could accelerate consensus protocols in peer-to-peer networks and decentralized systems.