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.