Path: utzoo!attcan!uunet!cs.utexas.edu!mailrus!ncar!boulder!bill
From: bill@boulder.Colorado.EDU
Newsgroups: comp.ai.neural-nets
Subject: Re: Parallel Processing/Neural Networks
Keywords: SIMD MIMD SYSTOLIC SCALING SIZE
Message-ID: <22042@boulder.Colorado.EDU>
Date: 8 Jun 90 17:37:48 GMT
References: <1576@gmuvax2.gmu.edu>
Sender: news@boulder.Colorado.EDU
Reply-To: bill@synapse.Colorado.EDU (Bill Skaggs)
Organization: University of Colorado, Boulder
Lines: 22

In article <1576@gmuvax2.gmu.edu> pmurali@gmuvax2.gmu.edu 
(Panchapagesan Murali) writes:
>I am interested in technical reports or good survey papers on "Highly
>Parallel Implementations" of contemporary neural net algorithms and
>an overview of the issues involved in every parallel implementation.
>
>In particular I am looking for SIMD/MIMD and Dataflow implementations
>and a classification that says which architecture is good for which
>algorithm and what is the scale of the neural nets in each
>implementation.  Can the scale of the neural net solve real life
>problems ? If so in what time span ?
>
  There are several papers dealing with these issues in the NIPS 89
volume.  (Full reference:  Advances in Neural Information Processing
Systems 2, ed. D Touretzky, Morgan Kaufman, San Mateo CA, 1990.)

  In particular, there is an article ("Computational Efficiency:  A
common organizing principle for parallel computer maps and brain
maps") discussing exactly the issues you raise.  There is also a
group of 10 papers dealing with hardware implementation.

	-- Bill Skaggs