Upgrade to Pro — share decks privately, control downloads, hide ads and more …

Periodic Patterns in Human Encounters

Matt J Williams
May 01, 2011
72

Periodic Patterns in Human Encounters

Research talk.
Venue: Research Retreat, Cardiff University School of Computer Science & Informatics.

Matt J Williams

May 01, 2011
Tweet

Transcript

  1. | Research Retreat | May 2011 |
    Matt Williams
    CS&I Research Retreat
    May 2011

    View full-size slide

  2. | Research Retreat | May 2011 |
    Periodic Patterns in Human
    Encounters
    Matt Williams
    CS&I Research Retreat
    May 2011

    View full-size slide

  3. | Research Retreat | May 2011 |
    Human encounters

    View full-size slide

  4. | Research Retreat | May 2011 |
    Why study human encounters?
    opportunistic content sharing
    virus spreading patterns

    View full-size slide

  5. | Research Retreat | May 2011 |
    Periodicity and patterns in human encounters

    View full-size slide

  6. | Research Retreat | May 2011 |
    • How prevalent are periodic encounter patterns in
    human networks?
    • How does the presence of periodic encounters
    affect information flow?
    • Can periodic patterns be detected and used to
    improve content sharing in opportunistic
    networks?
    Research questions

    View full-size slide

  7. | Research Retreat | May 2011 |
    How do we study human encounters?
    Smartphone Bluetooth
    encounters
    (@MIT)
    WiFi access point visits
    (@Dartmouth Campus)
    Foursquare venue visits
    (@Cardiff +others)

    View full-size slide

  8. | Research Retreat | May 2011 |
    How do we study human encounters?
    Smartphone Bluetooth
    encounters
    (@MIT)
    WiFi access point visits
    (@Dartmouth Campus)
    Foursquare venue visits
    (@Cardiff +others)

    View full-size slide

  9. | Research Retreat | May 2011 |
    • A group of nodes that regularly encounter one another
    with a given period
    Periodic encounter communities (PECs)

    View full-size slide

  10. | Research Retreat | May 2011 |
    • In opportunistic networks, a decentralised algorithm is
    needed
    Detecting PECs in OppNets

    View full-size slide

  11. | Research Retreat | May 2011 |
    Detecting PECs in OppNets

    View full-size slide

  12. | Research Retreat | May 2011 |
    • Period: the gap between
    reappearances of the community
    (24 hrs, 7 days, etc.)
    • Diameter: the distance between the
    most distant nodes in the community
    Properties of PECs
    diameter = 4

    View full-size slide

  13. | Research Retreat | May 2011 |

    View full-size slide

  14. | Research Retreat | May 2011 |
    • Diameter and period give us a theoretical
    worst-case for the time needed for all nodes to
    send messages to each other
    • In practice, how does information sharing
    compare to the worst-case?
    Information sharing

    View full-size slide

  15. | Research Retreat | May 2011 |
    in 75% of PECs,
    information sharing
    took less than 1/2 the
    worst-case time

    View full-size slide

  16. | Research Retreat | May 2011 |
    • PEC detection relies on a crisp definition
    • Inherent uncertainty in encounter times poses a
    problem
    Limitations

    View full-size slide

  17. | Research Retreat | May 2011 |
    • Dealing with fuzziness: borrow a technique from neuroscience!
    • Spike train synchrony: measures the similarity of bursting patterns
    of neurons
    Ongoing work: spike train methods
    Kreuz et al. 2009

    View full-size slide

  18. | Research Retreat | May 2011 |
    Human encounter ‘trains’
    Week 1
    Week 2
    Week 4
    Week 3

    View full-size slide

  19. | Research Retreat | May 2011 |
    Human encounter ‘trains’
    Week 1
    Week 2
    Week 4
    Week 3

    View full-size slide

  20. | Research Retreat | May 2011 |
    Regions of regularity
    Kreuz et al. 2009

    View full-size slide

  21. | Research Retreat | May 2011 |
    • Periodic encounter patterns exist and can be
    automatically detected
    • Evidence that periodic encounter patterns influence
    information flows in human encounter networks
    • Spike train methods are a promising solution for the
    detection of ‘fuzzy’ encounter patterns
    Summary
    Thanks for listening!

    View full-size slide

  22. | Research Retreat | May 2011 |
    References
    M.J. Williams, R.M. Whitaker, S.M. Allen, Decentralised detection of periodic
    encounter communities in opportunistic networks, Ad Hoc Networks,
    10.1016/j.adhoc.2011.07.008.
    T. Kreuz, D. Chicharro, R. G. Andrzejak, J. S. Haas, and H. D. I. Abarbanel,
    Measuring multiple spike train synchrony, Journal of Neuroscience
    Methods, vol. 183, no. 2, pp. 287–299, 2009.
    Attribution
    Library Courtyard. nevolution. http://www.flickr.com/photos/nevolution/
    2906377551/

    View full-size slide