Newsgroups: comp.ai.neural-nets
Path: cantaloupe.srv.cs.cmu.edu!das-news.harvard.edu!news2.near.net!MathWorks.Com!europa.eng.gtefsd.com!howland.reston.ans.net!cs.utexas.edu!utnut!wave.scar!93funkst
From: 93funkst@scar.utoronto.ca (FUNK  STEVEN LESLIE,,Student Account)
Subject: Re: C++ NN libraries
Message-ID: <CvGG8n.9oJ@wave.scar.utoronto.ca>
Sender: usenet@wave.scar.utoronto.ca
Nntp-Posting-Host: wave.scar.utoronto.ca
Reply-To: 93funkst@scar.utoronto.ca
Organization: University of Toronto - Scarborough College
References: <342ms3$g3j@lawelawe-f0.mrtc.maui.com>
Date: Thu, 1 Sep 1994 14:29:10 GMT
Lines: 13

Hi,

	I'm gonna go back and rewrite my connection object to see exactly how big it turned out.  BTW, I just realized that floats are 4 bytes not 2 bytes, its amazing what a little sleep can do.  Anywho, my point is that the object which I was using came out to at least 100 bytes.  This included the connection weight, the max, the min, a rest, a decay, and a mode.  At that point in time I was only able to model a system with a very few units and connections.  I switched over to a system where the weight alone is


 held in the matrix and all other information is held in another list and the systems performance increased dramatically.  
	Now, you can allocate as many blocks as you want but unless the blocks are contiquous or you have some way of handling them that makes for easy matrix manipulation then the problem still exists.  I could allocate more blocks too, but the largest single block was 64K.  I tried the global heap, but that gave me nothing but trouble.  Furthermore, I can allocate much more than 20 megs of real/virtual memory.  In fact there is no limit, except the ones imposed by the programmer, or the drive itself.  Because I


 use long int's to address the virtual memory, I can allocate a single contiguous block of up to (I think its) a trillion bytes.  This means that a matrix can hold a very large connection matrix, which is important in modelling 2-d layers.  Anywho, my point is that the objects have their place, and in a matrix doesn't seem to be it.

Steven

PS Sorry if this message gets choppy, but something between here and there has a tendency to mutilate things.

