BISC Seminar, 6 March 1997, 4-5:00pm, 310 Soda Hall

Michael Lee (leem@cs.berkeley.edu)
Fri, 28 Feb 1997 14:17:10 +0100


This is a multi-part message in MIME format.

--------------98115366643
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit

-- 
_______________________________________________________________________
Michael A. Lee
Berkeley Initiative in Soft Computing
387 Soda Hall                                      Tel: +1-510-642-9827
Computer Science Division                          Fax: +1-510-642-5775
University of California                    Email: leem@cs.berkeley.edu
Berkeley, CA 94720-1776 USA       WWW: http://www.cs.berkeley.edu/~leem
_______________________________________________________________________

--------------98115366643 Content-Type: text/plain; charset=us-ascii; name="zhong.3.6.97" Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="zhong.3.6.97"

A Hybrid Approach for Rule Discovery from Databases

BISC Seminar

Ning ZHONG Associate Professor, Ph.D. Dept. of Computer Science and Systems Engineering Faculty of Engineering, Yamaguchi University Tokiwa-Dai, 2557, Ube 755, Japan Tel&Fax: +81-836-35-9949 Email: zhong@ai.csse.yamaguchi-u.ac.jp

6 March 1997 310 Soda Hall 4:00-5:00pm

Abstract:

This talk introduces a new approach based on a hybrid intelligent model for rule discovery from databases. We first create an appropriate relationship between deductive reasoning and stochastic process, and extend the relationship for including abduction. Then, we define a Generalization Distribution Table (GDT), which is a variant of transition matrix in stochastic process, as a hypothesis search space for generalization, and describe that the GDT can be represented by connectionist networks. Since the creation of the connectionist networks is based on the GDT, the meaning of every unit in the networks can be explained clearly. Thus, not only the trained results in the connectionist networks are explicitly represented in the form of If-Then rule with Strength, but background knowledge can also be used for dynamically revising and changing the connectionist networks representation in a rule mining process. Hence, the connectionist networks are called "knowledge-oriented connectionist networks". Furthermore, we describe a rule mining process based on the connectionist networks representation. Finally, we introduce some extension for making our approach more useful, and discuss some issues about implementation of our approach.

The ultimate aim of the research project is to create an agent-oriented and knowledge-oriented hybrid intelligent model and system for knowledge discovery and data mining in an evolutionary, parallel-distributed cooperative mode. In this model and system, the typical methods of symbolic reasoning such as deduction, induction and abduction as well as the methods based on rough set and fuzzy set theory can be cooperatively used by taking the GDT, transition matrix in stochastic process, and their connectionist networks representation as the mediums. That is, the work that we are doing takes but one step toward this model and system.

---------------------------------------------------------------------------- CS-Seminars@CS.Berkeley.EDU

--------------98115366643--