Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!watmath!clyde!burl!ulysses!unc!mcnc!decvax!ittvax!dcdwest!sdcsvax!sdcrdcf!hplabs!sri-unix!HINTON@CMU-CS-C.ARPA
From: HINTON@CMU-CS-C.ARPA
Newsgroups: net.ai
Subject: Seminar - Learning in Production Systems
Message-ID: <468@sri-arpa.UUCP>
Date: Fri, 4-May-84 10:16:00 EDT
Article-I.D.: sri-arpa.468
Posted: Fri May  4 10:16:00 1984
Date-Received: Sat, 12-May-84 09:30:24 EDT
Lines: 23

From:  Geoff Hinton <HINTON@CMU-CS-C.ARPA>

          [Forwarded from the CMU-AI bboard by Laws@SRI-AI.]

The AI seminar on May 8 will be given by John Holland of the University
of Michigan.

Title: Learning Algorithms for Production Systems


Learning, broadly interpreted to include processes such as induction, offers
attractive possibilities for increasing the flexibility of rule-based
systems. However, this potential is likely to be realized only when the
rule-based systems are designed ab initio with learning in mind.
In particular, there are substantial advantages to be gained when
the rules are organized in terms of building blocks suitable for
manipulation by the learning algorithms (taking advantage of the
principles expounded by Newell & Simon).  This seminar will concentrate on:

  1. Ways of inducing useful building blocks and rules from experience,
     and
  2. Learning algorithms that can exploit these possibilities through
     "apportionment of credit" and "recombination" of building blocks.