NIIレクチャーシリーズのお知らせ
国立情報学研究所(NII)では、海外の情報学に関係する著名研究者を招聘し、レクチャーシリーズを行っております。 今回は、ＡＩの黎明期からＡＩの研究をされてきたカナダ・アルバータ大学のRandy Goebel先生の連続レクチャーのお知らせです。 今回はＡＩ原理をビッグデータに応用する最先端の研究の講演を行っていただきます。
講師:Prof. Randy Goebel (department of Computing Science at the University of Alberta, in Edmkonton, Alberta, Canada) 講義名 ：Do the emerging tools for managing big data fit with the founding principles of Artificial Intelligence? http://www.nii.ac.jp/en/event/list/0212 場所：国立情報学研究所 20階 2010室 講義日：2013/2/12, 20, 26, 3/1 時間: 13:30-15:00
出席は無料で、どなたでも参加できます。
参加のご検討よろしくお願いします。
佐藤 健 国立情報学研究所および総研大 ======= NII Lecture Series Title: Do the emerging tools for managing big data fit with the founding principles of Artificial Intelligence? ideas on the integration of the advice taker, structured inference, reasoning with incomplete information, and building multi-scale models from data.
Speaker: Prof. Randy Goebel (department of Computing Science at the University of Alberta, in Edmkonton, Alberta, Canada) He is also vice president of the innovates Centre of Research Excellence (iCORE) at Alberta Innovates Technology Futures (AITF), chair of the Alberta Innovates Academy, and principle investigator in the Alberta Innovates Centre for Machine Learning. He received the B.Sc. (Computer Science), M.Sc. (Computing Science), and Ph.D. (Computer Science) from the Universities of Regina, Alberta, and British Columbia, respectively. At AITF, Randy is in charge of reshaping research investments (graduate student scholarships, research chairs, research centres). His research interests include applications of machine learning to systems biology, visualization, and web mining, as well as work on natural language processing, web semantics, and belief revision. Randy has experience working on industrial research projects in crew scheduling, pipeline scheduling, and steel mill scheduling, as well as scheduling and optimization projects for the energy industry in Alberta. Randy has held appointments at the University of Waterloo, University of Tokyo, Multimedia University (Malaysia), Hokkaido University (Sapporo), and has had research collaborations with DFKI (German Research Centre for Artificial Intelligence), NICTA (National ICT Australia), RWC (Real World Computing project, Japan), ICOT (Institute for New Generation Computing, Japan), NII (National Institute for Informatics, Tokyo), and is actively involved in academic and industrial collaborative research projects in Canada, Australia, Europe, and China.
Abstract: The modern discipline of computer science has many facets, but what has clearly emerged in the last decade are three themes based on 1) rapidly accumulating volumes of data, 2) inter- and cross-disciplinary application of computer science to all scientific disciplines, and 3) a renewed interest in the semantics of complex information models, spanning a spectrum from semantic web, natural language, to multi-scale systems biology. This series of four lectures will attempt to knit together these three themes, by presenting the ideas that have emerged in their support: the rapid development and extension of machine learning theory and methods to help make sense of accumulating volumes of data, the application of computer science to nearly all scientific disciplines, especially those whose progress now necessarily relies on the management and interpretation of large data, and finally, the revival of a focus on semantics of information models based on data.
Outline: Lecture 1: Connecting Advice Taking and Big Data Lecture 2: Structured inference and incomplete information Lecture 3: Natural Language Processing: Compressing Data to Models Lecture 4: Hypothesis Management with Symbols and Pictures
Place: Lecture room 2010, 20th floor, National Institute of Informatics
Date: 13:30pm-15:00pm, February 12, 20, 26, March 1, 2013
Lecture 1 Connecting Advice Taking and Big Data Tuesday, February 12, 2013, 13:30 - 15:00 A fundamental premise of Artificial Intelligence (AI) is the ability for a computer program to improve its behaviour by taking advice. Incremental accumulation of advice or knowledge has never been easier than today, when the rate of data capture is higher than ever before, and the management of big data and deployment of machine learning are coupled to help manage the transition from data to knowledge. This lecture uses simple technical concepts from nearly sixty years of AI, to identify some of the research challenges of managing big data, and exploiting knowledge emergent from big data. The goal is to find some important research priorities based on the motivation of the Advice Taker, and the current state of big data management and machine learning.
Lecture 2 Structured inference and incomplete information Wednesday. February. 20, 13:30 - 15:00 If the foundation of Artificial Intelligence (AI) is the accumulation and use of knowledge, then a necessary stop is the structuring knowledge to be able to make inferences. The organization structures required to facilitate inference now span a broad spectrum of mathematical methods, including everything from simple propositional logic to sophisticated statistical and probabilistic inference. The two foundational components of computational inference are semantics of formal reasoning, and the development of reasoning methods to deal with incomplete information. This lecture reviews the foundational components of semantics and reasoning systems, including the development of goal-oriented reasoning based on abductive reasoning, the connection between logical and probabilistic systems, and especially how the architecture of reasoning systems can provide the basis for managing hypotheses in the face of incomplete information.
Lecture 3 Natural Language Processing: Compressing Data to Models Tuesday. February. 26, 13:30 - 15:00 The problem of machine processing of natural language (NLP) has long been a research focus of artificial intelligence. This is partly because the use of natural language is easily conceived as a cognitive task requiring human-like intelligence. It is also because the rational structures for computer interpretation of language require the full suite of computational tools developed over the last hundred years (grammar, dictionaries, logic, parsing, inference, and context management). Most of the recent practical advances in NLP have arisen in the context of simple machine learning applied to large language corpora, to induce fragments of language models that provide the basis for interpretive and generative manipulation of language. These largely statistical models are arisen in what has been called the "pendulum swing" of NLP, in which statistical models have recently dominated those based on structural linguistics. In this lecture, we look at the concept of noisy corpora and their role in language models, including some interesting alternative sources of data for building language models. The applications range from complex language summary to the information extraction from medical, legal, and historical documents.
Lecture 4 Hypothesis Management with Symbols and Pictures Friday. March. 1, 13:30 - 15:00 The current suite of Artificial Intelligence (AI) tools has provided a basis for sophisticated human-computer interfaces based on more than typing in language. In fact, one can develop multi-level representations that provide the basis for direct manipulation of visualizations. By constraining the repertoire of direct manipulations, one can enrich human computer interaction so that both humans and machines can understand and exploit visual interaction. This lecture shows how such direction manipulation requires a large repertoire of formal reasoning methods, and provides the sketch of formal framework and the problems arising in its development.