詳細資料
|
05/25(一)邀請林清詠教授來系演講
各位老師、同學： 系上曾新穆老師邀請 IBM T. J. Watson Research Center, New York, USA.的林清詠教授至系上演講，檢附詳細資訊及時間地點於文後，機會難得，請大家踴躍參加! 日 期：5月25日(一) 內容摘要： Social and biological networks have led to a huge interest in data analysis on graphs. Researchers within the KDD community have begun to study the task of data mining on graphs, including researchers from database-oriented graph mining, and researchers from kernel machine learning. We feel that exciting research problems and techniques can be discovered by exploring the link between these graph mining and kernel machine learning. This talk will present an overview of the techniques developed in graph mining and graph visualization. I will introduce the fundamental background in these fields and present exciting research problems at the interface of both fields. Moreover, I will introduce details in (1) Content-based network analysis: identify patterns from large content-based social networks; and (2) Context-sensitive graph visualization: present the patterns in the network and content modes for intuitive data exploration. These two parts are naturally complementary to each other. In particular, the patterns from network analysis provide a core structure for hierarchical and context-sensitive graph visualization. Likewise, the graph visualization helps better convey the patterns to users through an intuitive presentation. We analyze the content-based network tensors using a two level process. In this section we provide an overview of the overall process, and introduce some necessary notation. First, given a content-based network tensor, how to analyze the content and network modes in an efficient and robust manner? In this case, we need to model each mode of the tensor. We focus on clustering applications. Second, once the clusters on all modes are identified, how to efficiently identify the correlations across different modes? In this case, we need to correlate those clusters for crossmode patterns. For example, in the email tensor, we want to associate groups of people (communities) and groups of keywords (concepts) to find patterns of the form “who talks to whom about what.” In terms of visualization, we will introduce a high-order dimensionality reduction process using tensor decomposion, and leverage the low-dimensional representation along each mode for clustering. We utilize input tensor to find dense connections across clusters from different modes. 個人簡介： Dr. Ching-Yung Lin is a Research Scientist in IBM T. J. Watson Research Center, New York.He is currently leading projects on Collective Intelligence and Network Science researches. He is also an affiliate faculty at the University of Washington since 2003 and Columbia University since 2005. He received his B.S. and M.S. from National Taiwan University, in 1991 and 1993, respectively, and Ph.D. from Columbia University in 2000, all in electrical engineering. His research interest is mainly focused on multimodality signal analysis and complex network analysis, with applications on machine learning, distributed computing, embedded vision system, social computing and security. Dr. Lin invented and leads the Small Blue project, an IBM effort to build up a large-scale network analysis and visualization platform for utilizing collective intelligence and tacit knowledge mining. The goal is to develop advanced systems for search and recommendation on expertise, documents, webpages, software, communities, collaborators, etc, based on social, content and behavior analysis. By early April 2008, SmallBlue has been modeling the network, expertise and interest of more than 280,000 IBMers in more than 70 countries. An external commercial version of SmallBlue, called IBM Atlas (a.k.a. Lotus Atlas), was announced in Dec 2007 as the first enterprise social network analysis and expertise search engine in industry. In 2003, Dr. Lin created and led more than 100 researchers in 23 worldwide research institutes for the first large-scale collaborative video semantic annotation project. He pioneered the design of a semantic filtering framework which detects more than 100 visual concepts in videos. Dr. Lin's multimedia semantic mining project team performed best in the annual US National Institute of Standards and Technology (NIST) semantic video concept detection benchmarking 2002-2004. He also pioneered the design of video/image content authentication systems and a watermarking system surviving print-and-scan process. He has been serving as panelist, technical committee member, and invited speaker at various IEEE/ACM/SPIE conferences. He will be the general chair of IEEE International Conference on Multimedia and Expos (ICME) 2009 and the chair of the Circuits and Systems Multimedia Technical Committee 2010-2011. He is the Editor of the Interactive Magazines (EIM) of the IEEE Communications Society, 2004-2006, an associate editor of the IEEE Trans. on Multimedia 2004-2007, an editorial board member of Journal of Visual Communication and Image Representaiton, a guest editor of the Proceedings of the IEEE, Special Issue on Multimedia Security, June 2004, a guest editor of the EURASIP Journal of Applied Signal Processing, Special Issues on Visual Sensor Networks, Sept. 2006, and a Technical Program co-chair of IEEE ITRE 2003. Dr. Lin (co-)organized special sessions in the fields of multimedia security and multimedia understanding in IEEE ITCC 2001, ICIP 2003, ICME 2004 and ICIP 2004. He is a tutorial lecturer in ISCAS 2003, ICME 2003, Globecom 2003, ICCE 2005, ISCAS 2005, and MMSP 2005. Dr. Lin is a recipient of 2003 IEEE Circuits and Systems Society Outstanding Young Author Award, IBM Invention Achievement Awards in 2001, 2003 and 2007, IBM Research Division Award 2005, Acer Best EECS Thesis Award in 1993, and the Outstanding Paper Award in CVGIP 1993. He is the (co-)author of 120 journal articles, conference papers, book chapters and public release software. Dr. Lin holds four US patents and twelve pending patents. He is also a Ph.D. thesis co-advisor at National Taiwan University. Dr. Lin is a Senior Member of the IEEE and a member of ACM and INSNA.
相關網址：無 |