NetNews Usenet Archive 1992 #27/NN_1992_27.iso/spool/comp/ai/4333

Internet Message Format | 1992-11-18 | 3.7 KB

Xref: sparky comp.ai:4333 aus.ai:171 sci.math:15173 sci.math.stat:2373 aus.stats.s:898 alt.info-theory:64 comp.ai.philosophy:6837 sci.philosophy.tech:4162
Newsgroups: comp.ai,aus.ai,sci.math,sci.math.stat,aus.stats.s,alt.info-theory,comp.ai.philosophy,sci.philosophy.language,sci.philosophy.tech,melb.seminars,monash.seminars
Path: sparky!uunet!math.fu-berlin.de!informatik.tu-muenchen.de!regent!monu1.cc.monash.edu.au!monu6!bruce.cs.monash.edu.au!dld
From: [e-mail] (David L Dowe)
Subject: Monash seminar:  A model of inductive inference
Message-ID: <[e-mail]>
Summary: C S Wallace speaks on "A model of inductive inference"
Keywords: induction, inductive, inference, AI, artificial intelligence, information, information theory, estimation, prediction, Kolmogorov, complexity, Wallace
Sender: [e-mail] (USENET News System)
Organization: Computer Science, Monash University, Australia
Distribution: aus
Date: Wed, 18 Nov 1992 07:00:26 GMT
Lines: 54


                        DEPARTMENT OF COMPUTER SCIENCE
                AUSTRALIAN ARTIFICIAL INTELLIGENCE INSTITUTE

Next Seminar:  Wednesday, 25th November, 1992 4.15p.m.

Location:      Room 135 Dept of Computer Science, Monash Univ., Clayton

Topic:         A MODEL OF INDUCTIVE INFERENCE

Speaker:       Prof. C.S. Wallace, Dept of Computer Science, Monash

Abstract:

    Many problems in A.I., as in the real world, involve trying to
reach general conclusions from incomplete and often noisy data: the classic
problem of Inductive or 'Scientific' inference. Statistical inference can
also be included as a specially simple case.
    In most Logics, whatever conclusions are reached are (usually)
provably correct given the assumptions and data. In Induction, conclusions
are never provable, and are usually wrong. We first discuss why we bother
at all with such an unsatisfactory business, and in so doing distinguish
Induction from Prediction.
    Despite its importance as a mode of reasoning in scientific
enquiry and everyday life, the theoretical stucture of Induction remains
in dispute. Attempts such as Popper's and Hacking's to provide a LOGICAL
basis for induction seem to have fatal flaws, and even the limited case of
Statistical Induction presents uresolved questions.
    We offer a model of Induction having roots in Information Theory,
and connexions with Kolmogorov Complexity, Bayesian inference and formal
language theory. The approach presented will emphasize the formal
grammar view of the model and will expect little or no prior knowlege of
Information theory.
    The model seems to give a reasonably good account of Inductive
Inference, and has led to some quite powerful algorithms in the Machine
Learning field, which is the A.I. equivalent of Induction.
------------------------------------------------ End of Abstract ---------

   For those not familiar with the Monash (Clayton) grounds and/or wishing to
park here (at the Clayton campus), a map of the university grounds is in the
Melways street directory and can also be obtained from the University gatehouse
(off Wellington Rd).  An automatic ticket vending machine on the Western
stretch of the Ring Road sells daily parking permits for $0.60c .   The Dept of
Computer Science is located in Bldg 26, about 30 metres south across the lawn
from the Hargrave library.
                                        - - - - - - - - - - - - - - - - -

P.S.:  I post this not as the seminar co-ordinator but as one who believes the
talk could be of great interest to a widely-ranging audience.

Thank you and Yours (collectively) faithfully,   David Dowe.

(Dr.) David Dowe, Dept of Computer Science, Monash University, Clayton,
Victoria 3168, Australia  [e-mail]  Fax.:[phone]