Generalforsamling i Dansk Selskab for Teoretisk Statistik

Dagsorden for generalforsamlingen er udsendt per email til foreningens medlemmer.

I forlængelse af generalforsamlingen afholdes følgende foredrag:

SPEAKER: Anders Stockmarr (DTU Compute)

TITLE: Probabilistic Gossip

ABSTRACT: In an information network, information spreads through contact between nodes. I will picture the nodes as t.ex. bored housewives that exchange gossip; but it could also be servers exchanging information or similar. A contact between nodes can similar be perceived as a 'call'. Each agent possesses a bit of information, a 'secret'. When agents make contact, they exchange their information. Of interest is the time that passes until the network is saturated, ie. all agents knows all secrets. When calls are made at random, the process of known secrets is a Markov process with matrix-type state space. However, the possible number of calls are piling up as the number of agents increases, and to derive an algorithm for the distribution of the time to all secrets being known by all agents, it is advantageous to reduce the state space. Here the nature of the random selection of calls enters naturally. The presented work is unfinished analysis, and many problems remain. Application areas includes Artificial Intelligence.

The work is joint with Hans van Ditmarsch and Ioannis Kokkinis, University of Lorraine.

Reference: van Ditmarsch H., Kokkinis I., Stockmarr A. (2017) Reachability and Expectation in Gossiping. In: An B., Bazzan A., Leite J., Villata S., van der Torre L. (eds) PRIMA 2017: Principles and Practice of Multi-Agent Systems. Lecture Notes in Computer Science, vol 10621, Springer.