[logic-ml] NIIレクチャーシリーズのお知らせ(Randy Goebel)

[Ken Satoh]

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.