GCOE集中講義「数学と自然科学・社会科学 III 」 (2010年10月25日〜26日)

多様な人材育成プログラムの一環として、GCOE集中講義「数学と自然科学・社会科学 III 」を開講します。


ポスター(PDF) 文献表

日 程 10月25日(月)・26日(火) のそれぞれ 15:00〜17:00
講義名 数学と自然科学・社会科学 III
講 師 Mei Kobayashi 氏 (IBM Research-Tokyo)
場 所 京都大学理学部3号館108号室
タイトル From Information Retrieval to Web Mining: a mathematical tour
Part I: Overview
Part II: Selected Topics


    When I was a graduate student, Linear Algebra and Matrix Computations was my favorite course, because it was a beautiful bag filled with magical tricks. Many years down the road, while working in industry, I was delighted to find that the theorems, fast algorithms, and methods for minimizing error are not only being put to good use by search engines, but constitute the essential core of their service. Of course, no one knows all of the details about all search engines since so much of the knowledge is proprietary, but some basic algorithms that lie at the heart of some famous search systems have been published.  
    The first of this two-part lecture will be an overview of methods that led to the development of Web search and mining systems. The review will include the evolution of information retrieval systems (from static to dynamic databases), topic detection and tracking, clustering, and some basic algorithms used in Web search engines that exploit link structures. A selection of interesting, emerging topics will be introduced in the second lecture. Topics will include: On-line Discussion Mining, Blog Mining, Web Community Mining, and Recommendation Systems. Some algorithms developed at IBM Research (Almaden and Tokyo Laboratories) will be presented.    

Note: This two-part lecture will be accessible to any mathematics or science major with limited background in Web search and mining algorithms. Knowledge of matrix computations will be helpful. Computer science majors familiar with Web-related technologies may want to skip the first lecture.


    Dr. Mei Kobayashi graduated from Princeton University with an A.B. in Chemistry and the University of California at Berkeley with an M.A. and Ph.D. in Pure and Applied Mathematics. She worked at Lawrence Berkeley Laboratories as an undergraduate and graduate student intern in the Biochemistry and Math/Physics Divisions and as a Teaching Assistant/Grader for the Chemistry, Astronomy, and Mathematics Departments at Harvard University Graduate School of Arts & Sciences and the University of California at Berkeley.

    Dr. Kobayashi has been a Researcher at IBM for over 20 years, during which she worked on developing algorithms to solve inverse problems, simulations in low-end DASDs, wavelet analysis for speech synthesis, data hiding, information retrieval, data mining, and web log analysis. Recently, she was a Technical Assistant to the Director. While at IBM, Dr. Kobayashi taught courses on Matrix Computation and Wavelets at Tsukuba University, Hiroshima University and University of Electro-Communications and was a Visiting Professor at the Graduate School of Mathematical Sciences at the University of Tokyo for three years. Dr. Kobayashi's research interests include computational chemistry, numerical analysis, matrix computations, signal processing, wavelets, information retrieval and data mining. She has served on Technical Program Committees of SIAM Data Mining, ICPR and KDD and is a Member of the Editorial Board of the Communications of the ACM.