Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!zaphod.mps.ohio-state.edu!wuarchive!emory!hubcap!fpst From: shaun@isr.recruit.co.jp (Shaun Lawson) Newsgroups: comp.parallel Subject: parallel neural-net backpropagation Message-ID: <1991Apr24.120614.3729@hubcap.clemson.edu> Date: 24 Apr 91 09:00:30 GMT Sender: fpst@hubcap.clemson.edu (Steve Stevenson) Organization: Clemson University Lines: 34 Approved: parallel@hubcap.clemson.edu Return-Path: You may be interested in the following paper: T. Ito and K. Fukushima, Realization of a Neural Network Model Neocognitron on a Hypercube Parallel Computer, in the International Journal of High Speed Computing, Vol. 2, No. 1(1990)1-16. Here at ISR I use the NEC SX-2A supercomputer. I have a simple system which generates a highly vectorizable simulator program in FORTAN based on a user defined network definition. Learning takes place on the supercomputer, and after FTP'ing the weights file, the same program is used to run the network on a SPARCstation. I am interested in how you achieve the speedup of 7, with 8 SPARCs. Have you measured the throughput in terms of CUPS (Connections Updated Per Second)? If so, I would like to know what kind of performance you have achieved for what sized network(s). Best wishes, Shaun ------------------------------------------------------------------------------- Shaun Lawson Tel : (03)3536-7770 Institute for Supercomputing Research Fax : (03)3536-7769 1-13-1 Kachidoki, Chuo-ku Internet : shaun@isr.recruit.co.jp Tokyo, Japan 104 Disclaimers : Standard -- =========================== MODERATOR ============================== Steve Stevenson {steve,fpst}@hubcap.clemson.edu Department of Computer Science, comp.parallel Clemson University, Clemson, SC 29634-1906 (803)656-5880.mabell