Newsgroups: comp.ai.neural-nets
From: jimmy@ecowar.demon.co.uk (Jimmy Shadbolt)
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!news2.near.net!news.mathworks.com!usenet.eel.ufl.edu!usenet.cis.ufl.edu!caen!math.ohio-state.edu!howland.reston.ans.net!news.sprintlink.net!demon!ecowar.demon.co.uk!jimmy
Subject: Re: Generalization Improvement 
Distribution: world
References: <22OCT199418061165@summa.tamu.edu>
Organization: Econostat
Reply-To: jimmy@ecowar.demon.co.uk
X-Newsreader: Simple NEWS 1.90 (ka9q DIS 1.21)
Lines: 20
Date: Wed, 26 Oct 1994 13:45:20 +0000
Message-ID: <783179120snz@ecowar.demon.co.uk>
Sender: usenet@demon.co.uk

In article <22OCT199418061165@summa.tamu.edu> e0o0195@summa.tamu.edu writes:

>
>Hi,
>
>I wanted to know, if any one has found any correlation between the selection 
>of the initial weight space, and the generalization capability of the
>neural networks when trained for univariate or multivariate time series 
>datset.
>
>For example, the population in a city usually has an increasing trend.
                                                        ^^^^^^^^^^^^^^
>In scaling this data, between (0.0,1) the forecasting portion of data will 

        Trend!? Don't you de-trend your data to get N(0,1) distributed
        data that satisfy regression requirements?


        Drago.

