Path: utzoo!attcan!utgpu!jarvis.csri.toronto.edu!mailrus!tut.cis.ohio-state.edu!ukma!rutgers!mcnc!duke!romeo!hsg
From: hsg@romeo.cs.duke.edu (Henry Greenside)
Newsgroups: comp.ai.neural-nets
Subject: Re: Neural Net Applications in Chemistry
Summary: Can't forecast high-dimensional data...
Message-ID: <14480@duke.cs.duke.edu>
Date: 12 May 89 03:15:49 GMT
References: <1989May10.095408.5836@gpu.utcs.utoronto.ca> <TED.89May10130744@kythera.nmsu.edu> <201@bach.nsc.com> <39817@bbn.COM> <8382@phoenix.Princeton.EDU>
Sender: news@duke.cs.duke.edu
Reply-To: hsg@romeo.UUCP (Henry Greenside)
Distribution: na
Organization: Duke University CS Dept.; Durham, NC
Lines: 12

In these discussions of Farmer et al's methods versus
neural nets, has anyone addressed the real issue, how
to treat high-dimensional data?

In his paper, Farmer et al point out the crucial fact
that one can learn only low dimensional chaotic systems
(where low is rather vague, say of dimension less than
about 5). High dimensional systems require huge amounts
of data for learning. Presumably many interesting data
sets (weather, stock markets, chemical patterns, etc.)
are not low-dimensional and neither method will be
useful.