Sixth Workshop on Learning with Logics
and Logics for Learning




CALL for PAPERS and PARTICIPATION


July 6(Mon)-7(Tue), 2009
Kyodai Kaikan, Kyoto, Japan

Scope of the Workshop

Logic is one of the mathematical methods of representing data as well as rules in Machine Learning and Knowledge Discovery. Recently some methods are developed based on algebraic concepts, e.g. closed sets, for the aim. In the converse, Machine Learning procedures are found to provide procedural semantics to algebraic, and logical inference.

The aim of LLLL is to bring together researchers who are interested in both of the areas of machine learning and computational logic, and to have intensive discussions on various relations between the two with making their interchange more active. The LLLL workshop was started as a domestic workshop in 2002. As an international workshop, we held LLLL at Kitakyushu in 2005, at Tokyo in 2006, and at Miyazaki in 2007. LLLL2009 is supported by SIG-FPAI. LLLL2009 is collocated with DMSS2009, which is held July 7-8 at the same site. The CFP of DMSS2009 is available at
http://ibisml.org/dmss2009/


Potential (but not exclusive) topics include :

• Learning and knowledge discovery using logics
• Learning and knowledge discovery using algebraic methods
• Algorithmic aspects of learning based on logics
• Logics for machine learning and knowledge discovery
• Logics using machine learning
• Machine learning as a foundation of mathematics/mathematical procedures
• Amalgamation of logic-based learning and statistical/information theoretical learning
• Learning and knowledge discovery from relational data
• Learning and knowledge discovery from structured/semi-structured data

Registration fee: Free, but joining Banquet costs 6,000 JPY.

Inivited Speakers

Marta Arias Vicente (Universitat Polit鐵nica de Catalunya, Spain)

A Canonical Representation for Propositional Horn Theories and its Relation to Query Learning

Abstract: In this talk I will describe a canonical representation for propositional Horn theories known as the Guigues-Duquenne basis (or GD basis), that was introduced in the field of Formal Concept Analysis. I will show strong connections between this representation and two topics in query learning theory. First, I will show that a well-known algorithm by Angluin, Frazier and Pitt that learns Horn CNF always outputs the GD basis independently of the counterexamples it receives; second, and if time permits, I will show how to construct strong polynomial certificates from the GD basis directly. This is joint work with Jos・L. Balc痙ar.

Thomas Zeugmann (Hokkaido University)

The Normalized Information Distance and its Applications Using Graph Cuts

Abstract: First we recall some background necessary to understand the definition of the Normalized Information Distance. Then we outline as possibleapproximations the Normalized Compression Distance and the Google Distance. Next, we turn our attention to clustering. Clustering algorithms working with a matrix of pairwise similarities (kernel matrix) for the data are widely known and used, a particularly popular class being spectral clustering algorithms. In contrast, working directly with the pairwise distance matrix has found less attention despite the fact that in many applications distance matrices are often directly given. So, we look at clustering algorithms based on Semidefinite Programming (SDP) based on the work of Frieze and Jerrum and on spectral clustering working over similarity matrices. Finally, we shortly propose a simple heuristic for dealing with missing data, i.e., the case where some of the pairwise distances or similarities are not known.

Workshop Banquet

Workshop Banquet is held with DMSS at 18:00 on July 7th. Attendees who would like to join the banquet must send an email with the followings in the body:
- your name,
- affiliation, and
- email
Banquet fee is 6,000 JPY (JPY cash only).

How to Submit

LLLL2009 accepts extended abstracts written in English only. The abstracts are distributed in the workshop sites only and not considered to be formal publications. The extended abstract should be 1-8 pages in the format at the address:
http://www.dumbo.ai.kyutech.ac.jp/LLLL/jsaiws_template.lzh

The authors should send their submitted papers as a PDF file to
akihiro@i.kyoto-u.ac.jp,
hirata@ai.kyutech.ac.jp, and
minato@ist.hokudai.ac.jp (all).

Please write the text `[LLLL09: Paper Submission]' in the subject of the massage, and put the following in the body:
- Title of your paper,
- A list of all authors and their affiliations (please add "*" mark to the presenter),
- Name, Telephone number, and E-mail address of the corresponding author.

Meeting Venue

Kyodai-Kaikan Tel: +81-75-751-8311 Fax: +81-75-761-5403 URL : http://www.kyodaikaikan.jp/ Access: - From Kyoto JR Station, take the bus 206 at terminal D2, stop at the bus stop Kyodai-seimon-mae and walk about 5 mins. - By Keihan train, get off at Keihan Marutamachi Station and walk about 7 mins.

Committee Members

Workshop Chair

Akihiro Yamamoto (Kyoto University)

PC Co-chairs

Kouichi Hirata (Kyushu Institute ofTechnology)
Shin-ichi Minato (Hokkaido University)

Program Commitee (current status)

Yoji Akama (Tohoku University)
Marta Arias (Universitat Politecnica de Catalunya)
Kouichi Hirata (Kyushu Institute ofTechnology)
Tamas Horvath (University of Bonn and Fraunhofer IAIS)
Katsumi Inoue (National Institute of Infomatics)
Shin-ichi Minato (Hokkaido University)
Tetsuhiro Miyahara (Hiroshima City University)
Taisuke Sato (Tokyo Institute ofTechnology)
Takayoshi Shoudai (Kyushu University)
Hiroo Tokunaga (Tokyo Metropolitan University)
Gyorgy Turan (University of Illinois)
Akihiro Yamamoto (Kyoto University)

Website and Contact

http://www.iip.ist.i.kyoto-u.ac.jp/LLLL09/

Postal addess : Akihiro Yamamoto