Newsgroups: comp.ai.neural-nets
Path: cantaloupe.srv.cs.cmu.edu!rochester!udel!news.mathworks.com!usenet.eel.ufl.edu!europa.chnt.gtegsc.com!howland.reston.ans.net!tank.news.pipex.net!pipex!news.sprintlink.net!redstone.interpath.net!sas!mozart.unx.sas.com!saswss
From: saswss@hotellng.unx.sas.com (Warren Sarle)
Subject: Re: Minsky's Interacting Causes
Originator: saswss@hotellng.unx.sas.com
Sender: news@unx.sas.com (Noter of Newsworthy Events)
Message-ID: <DCAB29.339@unx.sas.com>
Date: Tue, 25 Jul 1995 18:22:57 GMT
X-Nntp-Posting-Host: hotellng.unx.sas.com
References:  <push-2207950118170001@mind.mit.edu>
Organization: SAS Institute Inc.
Lines: 47


In article <push-2207950118170001@mind.mit.edu>, push@mit.edu (Pushpinder Singh) writes:
|> Marvin Minsky once drew a diagram that I liked very much.  It went
|> something like this:
|>
|>                       Interacting Causes
|>
|>               few <-- number of causes --> many
|>           +----------------------------------------+
|>    /|\    | Easy        Linear,    Neural Networks |
|>     |     |           Statistical    Fuzzy Logic   |
|>     |     |                                        |
|>  scale of |    Qualitative           Analogical    |
|>   effect  |     Reasoning            Reasoning     |
|>     |     |                                        |
|>     |     |  Logical    Case Based     Intractable |
|>    \|/    | Reasoning   Reasoning           \      |
|>           +----------------------------------\-----+
|>                                               \
|>                                   Find better representation!

I take it that "cause" is intended to mean "input" in the neural net 
sense, although that is a major blunder in philosophy of science.

What does "scale of effect" mean?

|> Of course Minsky is the first to admit this is an oversimplification,
|> but it is a step towards a theory of how the pieces fit together.
|> More specifically, it offers a theory about how to decide what
|> approach to use when dealing with problems that involve change.

What does "change" have to do with the diagram?

|> What I would like to raise is, what's wrong with this diagram?

If the diagram is saying that linear statistical models are suitable
for a moderate number of inputs while neural nets are suitable for
many inputs, then it is quite wrong, since linearity and number of
inputs are separate issues. Also, the statistical varieties of
case-based reasoning are nonlinear, so the diagram seems to
incorporate nonlinearity on both axes.

-- 

Warren S. Sarle       SAS Institute Inc.   The opinions expressed here
saswss@unx.sas.com    SAS Campus Drive     are mine and not necessarily
(919) 677-8000        Cary, NC 27513, USA  those of SAS Institute.
