Newsgroups: comp.ai.neural-nets
Path: cantaloupe.srv.cs.cmu.edu!rochester!udel!gatech!howland.reston.ans.net!news.sprintlink.net!redstone.interpath.net!sas!mozart.unx.sas.com!saswss
From: saswss@hotellng.unx.sas.com (Warren Sarle)
Subject: Re: Problem...
Originator: saswss@hotellng.unx.sas.com
Sender: news@unx.sas.com (Noter of Newsworthy Events)
Message-ID: <D7yz4G.3uv@unx.sas.com>
Date: Tue, 2 May 1995 21:17:04 GMT
X-Nntp-Posting-Host: hotellng.unx.sas.com
References:  <jpullen.1.0000F249@arn.net>
Organization: SAS Institute Inc.
Lines: 65


Everyone who uses neural nets should be required to take a year of
statistics first.

In article <jpullen.1.0000F249@arn.net>, jpullen@arn.net writes:
|> Hi there.  I have been experimenting with NNs for a few weeks now and have run
|> into some problems. ... I have been trying to predict elapsed times for a race car
|> for the quarter mile based on the weather conditions.  I have not had any luck
|> getting WINNN or NNDT to learn the patterns. ...  Here
|> is a sample of the data presented in the format of:
|>
|> input    output
|>
|> 5449 9.007
|> 6332 9.040
|> 5241 9.069
|> 5255 9.082
|> 4961 9.129
|> 5402 9.132
|> 5041 9.134
|> 5025 9.138

The first thing you do is look at a scatter plot:

          Plot of OUTPUT*INPUT.  Legend: A = 1 obs, B = 2 obs, etc.

 OUTPUT |
        |
   9.15 +
        |
        |            B
        |          A              A
        |
        |
        |
   9.10 +
        |
        |
        |                    A
        |                    A
        |
        |
   9.05 +
        |                                                           A
        |
        |
        |
        |
        |                           A
   9.00 +
        |
        ---+--------+--------+--------+--------+--------+--------+--------+--
         4750     5000     5250     5500     5750     6000     6250     6500

                                        INPUT

While the plot suggests that output may be a decreasing function of
input, it should be obvious that the data are inadequate to make
predictions via neural nets or anything else. Get more data.

-- 

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.
