Newsgroups: comp.ai.neural-nets
Path: cantaloupe.srv.cs.cmu.edu!rochester!cornellcs!newsstand.cit.cornell.edu!portc01.blue.aol.com!portc02.blue.aol.com!newsfeed.pitt.edu!news.duq.edu!newsgate.duke.edu!interpath!news.interpath.net!news.interpath.net!sas!newshost.unx.sas.com!saswss
From: saswss@hotellng.unx.sas.com (Warren Sarle)
Subject: Re: missing data
Originator: saswss@hotellng.unx.sas.com
Sender: news@unx.sas.com (Noter of Newsworthy Events)
Message-ID: <E1Ls3A.Mu3@unx.sas.com>
Date: Thu, 28 Nov 1996 23:04:22 GMT
X-Nntp-Posting-Host: hotellng.unx.sas.com
References:  <3289990F.5206@tamaya.avignon.inra.fr>
Organization: SAS Institute Inc.
Lines: 34


In article <3289990F.5206@tamaya.avignon.inra.fr>, Marie WEISS <weiss@tamaya.avignon.inra.fr> writes:
|> I assume that I have trained my neural network with 15 data as inputs.
|> Now, I want to test it on other data, but, unfortunately, I have the
|> values of only 13 data of the 15 needed.
|> Is there a solution to this problem ?
|> 
|> ------------------------------------------------------------------------
|> Marie WEISS                         
|> Bioclimatologie                     
|> INRA - Avignon
|> France                             E-mail: weiss@tamaya.avignon.inra.fr
|> ------------------------------------------------------------------------

Quick and dirty: Replace the two missing input values by the means or
medians of those two input variables.

Slower and better: Compute the conditional distribution of the two
missing inputs given the 13 non-missing inputs. Find the expected
value of the network output by integrating over this conditional
distribution. See, for example:

Tresp, V., Neuneier, R., and Ahmad, S. (1995), "Efficient methods for 
dealing with missing data in supervised learning", in Tesauro, G.,
Touretzky, D.S., and Leen, T.K., (eds.) Advances in Neural Information 
Processing Systems 7, Cambridge, MA: The MIT Press, pp. 689-696.

-- 

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.
 *** Do not send me unsolicited commercial or political email! ***

