Newsgroups: comp.ai.neural-nets
Path: cantaloupe.srv.cs.cmu.edu!rochester!udel!news.mathworks.com!hookup!news.moneng.mei.com!uwm.edu!reuter.cse.ogi.edu!henson!news
From: mobus@sanjuan.cs.wwu.edu (George Mobus)
Subject: Re: Time based neural nets
Message-ID: <1995Feb12.230056.3295@henson.cc.wwu.edu>
Sender: news@henson.cc.wwu.edu (USENET-WWU)
Organization: Western Washington University
References: <Pine.SOL.3.91.950209113259.8959K-100000@bingsun1>
Date: Sun, 12 Feb 1995 23:00:56 GMT
Lines: 38

In article <Pine.SOL.3.91.950209113259.8959K-100000@bingsun1> Scott  
Hackett <br00372@bingsuns.cc.binghamton.edu> writes:
> I am working on my thesis, creating a temporal neural network.  There is 
> so little literature out there that I am forced to play trial and error 
> with the design.  Does ANYBODY have any knowledge, references or 
> suggestions about neural nets that process time based information?  My 
> inputs are not just vectors, but vector series over time.  Any help  
would 
> be very greatly appreciated.
> 
> Scott Hackett
> br00372@bingsuns.cc.binghamton.edu
>  

As some other posters have mentioned, there is quite a rich literature on  
time-series and sequence learning/generating.  For a different view of the  
role of time in neuromimic architectures, I'd like to suggest my own work  
as reported in Chapter 13, Toward a Theory of Learning and Representing  
Causal Inferences in Neural Networks in "Neural Nets for Knowledge  
Representation and Inference", Levine and Aparicio (Eds.), Lawrence  
Erlbaum, Associates, 1994.  I develop a model that incorporates several  
time scales of memory trace and association encoding that accounts for the  
ability to retain long-term associations that may be over-ridden by  
short-term experiences.  If the latter are not reinforced over time,  
however, they decay and the original long-term memory reemerges to  
influence behavior.

Good luck in your efforts.

George

+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
George Mobus                    | Opinions expressed are my own and do
Adjunct Assistant Prof.         | not represent those of my employer or
Dept. of Computer Science       | any generally rational being.
Western Washington University   |
mobus@sanjuan.cs.wwu.edu        |
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
