ACM SIGMOD日本支部の皆様
日本データベース学会の皆様

WWWグラフの大御所、IBMのDr. Andrew Tomkinsの講演が
以下のとおりに行われます。奮ってご参加ください。

　　　　　　　　講演会のご案内
　　　　　　　主催　ACM SIGMOD日本支部
　　　　　　　協賛　日本データベース学会

日時　５月１６日（月）　午後４時〜５時００分
場所　東京大学生産技術研究所　会議室Ｂ　５階（Ew-502）
      E棟　5階　
      http://www.iis.u-tokyo.ac.jp/map/index.html

Title: Social Network Graphs: Visualization of Relationships,
       and Analysis of Geographic Connections
Speaker: Andrew Tomkins (IBM Almaden Research Center)

参加費　無料

参加ご希望の方は、ＳＩＧＭＯＤホームページにて
     （ http://www.sigmodj.org ）
会員登録の後（会費無料、すでに登録されている方は結構です）、
      sigmodj_lecture@tkl.iis.u-tokyo.ac.jpに
添付の参加申込書をお送り下さい。

皆様のご参加をお待ちしております。

                        ACM　SIGMOD日本支部　支部長　北川博之
　　　　　　　　　　　　　　　　　　　　　　 担当幹事　中野　美由紀

                        連絡(問合せ)先　ACM SIGMOD日本支部
                                sigmodj_lecture@tkl.iis.u-tokyo.ac.jp
　　　　　　　　　　　　　　　　http://www.sigmodj.org

-----------------------------------------------------------------
To: sigmodj_lecture@tkl.iis.u-tokyo.ac.jp

ACM SIGMOD日本支部　講演会　参加申し込み

5月16日（月）の講演会に参加
・名前　　　
・ご所属
------------------------------------------------------------------

Lecture: Dr. Andrew Tomkins (IBM Almaden Research Laboratory)

Title: Social Network Graphs: Visualization of Relationships, and
Analysis of Geographic Connections

Abstract:

In this talk, we will cover two pieces of work on social network
graphs.  The first covers the following problem: given a social
network graph and two individuals, present a small subgraph that
best shows the connection between the two individuals.  I will
argue that connections between individuals should be manifested,
not by paths, but by these carefully-chosen subgraphs, and I'll
give algorithms to generate these subgraphs in real-time on large
out-of-core graphs with over one billion edges.

Next, I'll turn to an analytical discussion of the connection between
friendship and geographic distance.  Earlier models suggest that the
probability of two individuals being friends should be proportional
to the inverse square of the distance between them if the graph is
to have short paths.  We examine a large real social network based on
the LiveJournal blog hosting site, and show that the graph contains
short paths but does not manifest inverse-square friendship probabilities.
This leads to an improved characterization of the probability of two
people becoming friends which takes into account the population density
of the environment in a natural way.  We give both experimental and
theoretical results regarding this new model.

------------------