11:00
12:00

Robin Vacus (Bocconi University of Milan)

Title: Information Spread with Minimal Resources: some insights from the bit-dissemination problem.

Abstract:

Multi-agent systems often face two closely related challenges: achieving consensus among agents (i.e., making sure they take consistent decisions), and aggregating local information about their environment. Indeed, reaching approximate agreement is a typical prerequisite of collective decision-making, while the overall performance of the group depends on how effectively local information is processed.

In this talk, I will present the /bit-dissemination problem/, an idealized model which attempts to capture these challenges. It involves a fully-connected network of n agents holding binary opinions, one of which is called “correct”
. The group’s objective is to achieve consensus on this correct opinion, which is initially known only to a single “
source” agent. I will present several protocols for solving this problem, focusing on simple, natural algorithms that require minimal memory, synchronization, and communication among agents. I will also discuss lower bound techniques, in order to shed light on the fundamental limitations of information spread in distributed systems.

https://algodist.labri.fr/index.php/Main/GT

LaBRI salle 178 - lien visio https://webconf.u-bordeaux.fr/b/arn-4tr-7gp