Newsgroups: comp.ai.neural-nets
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!news2.near.net!news.mathworks.com!news.alpha.net!uwm.edu!math.ohio-state.edu!jussieu.fr!centre.univ-orleans.fr!univ-lyon1.fr!swidir.switch.ch!newsfeed.ACO.net!Austria.EU.net!EU.net!uknet!festival!hwcee!dsr
From: dsr@cee.hw.ac.uk (Donald Reay)
Subject: Re: Non-linear plant with hysteresis (how to control using ANN)
Message-ID: <Cy0ovC.Dyv@cee.hw.ac.uk>
Sender: news@cee.hw.ac.uk (News Administrator)
Organization: Dept of Computing & Electrical Engineering, Heriot-Watt University, Scotland
References: <37k577$ote@perseus.ee.ubc.ca> <37ksrn$mm7@maui.cs.ucla.edu> <andersen.782528048@s1.elec.uq.oz.au>
Date: Fri, 21 Oct 1994 09:54:47 GMT
Lines: 31

In article <andersen.782528048@s1.elec.uq.oz.au> andersen@s1.elec.uq.oz.au (Hans Andersen) writes:
[snip]
>
>I would like to point out an alternative approach to neural control
>called "indirect learning".  In this method the error is not back-
>propagated through the plant.  It uses a different concept which can
>be described intuitively as follows:
>
>   Try something.  If it doesn't have the desired effect - don't
>   worry about how wrong it was, but remember what effect it did 
>   have so that you can use that knowledge in the future.
>
I too have used this training method to good effect. I first saw the
technique described by J.S. Albus

Albus, J.S.
Data Storage in the Cerebellar Model Articulation Controller (CMAC)
Trans. ASME, J. DSMC, 97(3), Sept. 1975

who referred to it as a "time inversion technique" and I have always
referred to it as Albus' time inversion training technique. My question
is: Is "indirect learning" a more technical term for this and where in
the literature was it first discussed?

Donald Reay
Department of Computing and Electrical Engineering
Heriot-Watt University
Edinburgh
UK


