neuron-request@hplabs.hp.COM (Neuron-Digest Moderator Peter Marvit) (10/03/88)
Neuron Digest Sunday, 2 Oct 1988 Volume 4 : Issue 10 Today's Topics: Re: Neuron Digest V4 #6 (proceedings) Rochester Connection Simulator information Re: Neuron Digest V4 #6 (proceedings) Re: Neuron Digest V4 #6 (proceedings) Re: Neuron Digest V4 #6 (pulse-coded networks) Analogical reasoning Re: Neural Nets??? BUMBLE-BEE! Send submissions, questions, mailing list maintenance and requests for back issues to "Neuron-request@hplabs.hp.com" ------------------------------------------------------------ Subject: Re: Neuron Digest V4 #6 From: brp@sim.uucp (bruce raoul parnas) Date: Tue, 27 Sep 88 02:30:51 +0000 In article <7753@burdvax.PRC.Unisys.COM> pastor@bigburd.PRC.Unisys.COM (Jon Pastor) writes: >the special issue containing the abstracts for INNS). There are no plans to >publish proceedings, and the reason is financial. INNS wished to keep the >cost of the conference down, so as to make it accessible to as many >researchers and students as possible. The INNS board decided that the >inclusion of proceedings would have increased the cost of conference >registration by an unacceptable amount (let's say, $90, based on the ICNN >Proceedings costs). I have had some dealings with the INNS, and I find that they are very concerned about money, to the exclusion of the interests of science. But that's another issue. The IEEE Neural and Information Processing Systems conference held last November in Denver DID have available proceedings which could be purchased by those who attended the conference for about $29 or so. The cost of registration was $50 (for students, and pretty low for others), which is considerably less than that of the INNS conference. They did not seem to have a problem making proceedings available. These were done (as you suggested) by getting the orders FIRST, then publishing. Still, the conference was cheaper, probably equally good, and there were no problems with proceedings. I think the problem lies more with the administration at INNS than it does in fact. - -bruce [[ As was pointed out in the last Digest, INNS is a young organization. I hope people will provide useful and constructive feedback to INNS board members and conference organizers. I would be dubious about inuendos without suuport which otherwise detract from good and valid criticisms. -PM]] ------------------------------ Subject: Rochester Connection Simulator information From: bukys@cs.rochester.edu Date: Tue, 27 Sep 88 09:57:20 -0400 The Rochester Connectionist Simulator is being made available by the University of Rochester for research purposes. No commercial use is allowed without the explicit written permission of the University of Rochester. =============================================================================== You can reach other users of the simulator via the users' mailing list: <simulator-users@cs.rochester.edu> If you are not on this mailing list, and wish to be added, send a note to <simulator-request@cs.rochester.edu> Please send bug reports to <simulator-bugs@cs.rochester.edu> We are interested in fixing bugs, but can't make any promises! Please make your bug reports as specific as possible. =============================================================================== If you are unable to obtain anonymous FTP access to the simulator distribution, you can still order a copy the old way. Contact Rose Peet Computer Science Department University of Rochester Rochester, NY 14627 (USA) or <connect@cs.rochester.edu>, and she will send you the approriate forms. We are currently charging $150 for a distribution tape and a manual. If you have TeX and a PostScript printer, you should be able to produce your own copy of the 181-page manual. If you want a paper copy of the manual anyway, send a check for $10 per manual (payable to the University of Rochester) to Rose Peet at the above address. We do not have the facilities for generating invoices, so payment is required with any order. =============================================================================== This directory (public/rcs on CS.Rochester.EDU) contains release 4.1 of the Rochester Connectionist Simulator. Version 4.1 was released on April 5, 1988, and can be found in the file "rcs_v4.1.tar.Z". It is a compressed tar file, size 837215 bytes. When uncompressed, it should be size 3092480 bytes, and running the Unix "sum" program on it should yield "62514 3020". REMEMBER TO USE THE "TYPE TENEX" or "TYPE BINARY" COMMAND IN FTP WHEN RETRIEVING COMPRESSED TAR FILES!!! Official bug reports also reside in this directory: rcs_v4.1.patch.01 fixes a SERIOUS problem in the logging feature. rcs_v4.1.patch.02 fixes some minor syntactic problems in the doc. rcs_v4.1.patch.03 a missing "return" makes commands noops on SUN4 The patches can be applied by hand if necessary, but you will make your life easier if you obtain the widely-available "patch" program, and redirect each patch file into "patch -p". Other files of interest: rcs_v4.1.note.01 the doc's missing "urcsmanual" TeX files. rcs_v4.1.doc.tar.Z the manual, in postscript format, ready to print =============================================================================== ------------------------------ Subject: Re: Neuron Digest V4 #6 From: tomh@proxftl.UUCP (Tom Holroyd) Date: 27 Sep 88 14:03:21 +0000 Well, as I understand it, the INNS (which had its first conference in Boston Sep. 6-10) is not associated with the IEEE. The San Diego conference WAS held by the IEEE, so you should be able to get proceedings from the above address. Copies of the INNS 1st Annual meeting ABSTRACTS can be obtained from: Pergamon Press, Inc. Journal Fulfillment Dept. Fairview Park Elmsford, NY 10523 USA More information about the INNS can be obtained from: Boston University Center for Adaptive Systems 111 Cummington Street Boston, MA 02215 USA Again, the IEEE NN conference and the INNS conference are *different* (this issue will be further confused by the fact that next year, the two conferences will be held in the same city, during the same month, to wit: Wasington, DC, Sep. 5-9, 1989, Omni Shoreham Hotel, for the INNS conference). For INNS conference info, try: INNS Conference J.R. Schuman Associates 316 Washington St. Box 125 Wellesley, MA 02181 USA (617)237-7931 Tom Holroyd UUCP: {uflorida,uunet}!novavax!proxftl!tomh [[ Thanks again for the correction and additional information. Now can I get some of the readers to proofread my papers as carefully as you note the Digest (*smile*). A side note -- I'm a bit concerned about the growing number of "Neural Net" conferences and organizations. I still contend that the field is too young to start fragmenting without drawing clear and *non-competitive* lines of distinction between groups. I welcome feedback on this somewhat extra-technical topic. -PM ]] ------------------------------ Subject: Re: Neuron Digest V4 #6 From: dhw@itivax.UUCP (David H. West) Date: 27 Sep 88 14:14:13 +0000 In article <7753@burdvax.PRC.Unisys.COM>, pastor@bigburd.PRC.Unisys.COM (Jon Pastor) writes: > There are no plans to > publish proceedings, and the reason is financial. INNS wished to keep the > cost of the conference down, so as to make it accessible to as many researchers > and students as possible. Ingenious. Many of us unable to attend for financial reasons were hoping to compensate by reading the proceedings. I expect that usually more people (only) read the proceedings of a conference (within 6 months, say) than attended it. > While making the proceedings available at an additional charge would seem to > have been a viable alternative, the economics of publishing are such that > this was ruled out [...] I believe we are constrained here by those secondary goals of academic publishing which have displaced the primary goal of getting the information out. The network technology that enables you to read this message, and to obtain data from archive sites, is capable of handling most of this problem. There are of course difficulties with the volume of the data, and with diagrams, and there are various ways round those difficulties. GNU (FSF) software is a fine example of how something useful but bulky can be widely available without formal publication. Are we collectively too stupid to apply this lesson to non-executable information? The major residual problem that I see is that this method marginalizes those without net access (or the right kind of net access). Again, it's hard to believe that we have insufficient ingenuity to solve this. - -David West dhw%iti@umix.cc.umich.edu {uunet,rutgers,ames}!umix!itivax!dhw CDSL, Industrial Technology Institute, PO Box 1485, Ann Arbor, MI 48106 [[ See my comments in the last issue about "on-demand" printing as well. I'm not always comfortable with holding FSF as a model, especially since the conditions are somewhat different for an ongoing piece of software and a one-time publication of proceedings. Nonetheless, the points above are well made and well taken. -PM ]] ------------------------------ Subject: Re: Neuron Digest V4 #6 (pulse-coded networks) From: pschmidt@LISPERANTO.BBN.COM Date: Tue, 27 Sep 88 13:48:40 -0400 In the last issue, Dan Mocsny asked about NNs that try to model the propagation of potential-spikes. These networks can be termed "pulse-code networks". Though I am not familiar with the paper he references, I recently completed my bachelor's thesis at MIT on the behavior of such a network model, and in the process I compiled a bibliography of some of the relevant work on such models. If anyone is interested, I'd be glad to supply a copy of my thesis - I've attached the abstract and bibliography below. Cheers, Peter H. Schmidt pschmidt@bbn.com C u t H e r e - ------------------------------------------------------------------------------ Title: The Transfer Characteristic of Neurons in a Pulse-Code Neural Network A pulse-code neural network differs from most contemporary, activation-based, neural networks in that it retains information they discard. The state of pulse-code neurons is contained in the values of their soma and hillock potentials. The outputs of a pulse-coded neuron are the actual firing events in time; this allows observation of meaningful rhythmic behavior of the type which has been seen in biological neural networks, but which is averaged away in activation-based networks. This thesis examines the transfer characteristic of a single pulse-code neuron with one inhibitory and one excitatory input. Data was obtained through simulation of two versions of the model which differ in the way the soma potential is updated after an input event occurs. In the first version, it is increased or decreased in fixed increments and simply hard limited at its extremes, while in the second it is increased or decreased fractionally towards asymptotic limits. Analytic expressions are developed to describe the time-evolution of the state variables of both versions; this analysis is supported by the data. Based on this analysis, the fractional version of the model is recommended as the one worthy of further investigation. (Sorry about the Scribe format commands...) @Article[ McCandP ,Key="McCandP" ,Author="McCulloch, Warren S. and Walter H. Pitts" ,Title="A Logical Calculus of the Ideas Immanent in Nervous Activity" ,Journal="The Bulletin of Mathematical Biophysics" ,Year="1943" ,Number="5:115-133" ] @InCollection[ NervousCoding ,Key="RaymondLettvin" ,Author="Raymond, S.A. and J.Y. Lettvin" ,Title="Aftereffects of Activity in Peripheral Axons as a Clue to Nervous Coding" ,BookTitle="Physiology and Pathobiology of Axons" ,BookAuth="Waxman, S.G., ed." ,Publisher="Raven Press" ,Year="1978" ] @Article[ EffNervImp ,Key="Raymond" ,Author="Raymond, Stephen A." ,Title="Effects of Nerve Impulses on Threshold of Frog Sciatic Nerve Fibres" ,Journal="Journal of Physiology" ,Year="1979" ] @Article[ MultipleMeaning ,Key="Chung" ,Author="Chung, Shin-Ho, S.A. Raymond, and J.Y. Lettvin" ,Title="Multiple Meaning in Single Visual Units" ,Journal="Brain, Behavior and Evolution" ,Year="1970" ] @Unpublished[ Pratt ,Key="Pratt" ,Author="Pratt, Gill" ,Title="RLE Progress Report" ,School="MIT Research Laboratory of Electronics" ,Month=apr ,Year="1986" ,Note="Report on Doctoral Research" ] @Unpublished[ Mathtroff ,Key="PrattSchmidt" ,Author="Pratt, Gill and Peter H. Schmidt" ,Title="Mathematics for Pulsed Neural Network Event Driven Simulation" ,School="Massachusetts Institute of Technology" ,Month=aug ,Year="1987" ] ------------------------------ Subject: Analogical reasoning From: rickel@cs.utexas.edu (Jeff Rickel) Date: Tue, 27 Sep 88 14:21:40 -0500 Has there been any work on neural network models for solving geometry analogy problems like those in Evans famous dissertation in the 60's? Anything similar? Jeff Rickel rickel@cs.utexas.edu ------------------------------ Subject: Re: Neural Nets??? BUMBLE-BEE! From: sbrunnoc@hawk.ulowell.edu (Sean Brunnock) Date: 27 Sep 88 20:32:00 +0000 In article <551@epicb.UUCP> david@epicb.UUCP (David P. Cook) writes: > What's the deal??? Neural nets are the wave of the future... I know > there are LOTS of people and research institutions working on them out > there and there are already several commercially available products. Having attended the INNS conference and trying out many of the exhibitor's products firsthand, I am not impressed with what is commercially available in terms of neural network products and programs. I believe that most of the products at the conference can be categorized into three groups: books, programs to allow people to build neural networks, and handwriting analyzers. Books, I cannot complain about. Neural net builders are nice but not groundbreaking. The handwriting analyzers did not fare well with me. These neural net programs didn't do any better than signature table products of twenty years ago. Of course the neural net programs can be retrained. But Nestor's handwriting analyzer, after five days of training by various attendees of the coference, gave me a response of "re11o" after I wrote "hello". My handwriting is legible, and I was not trying to fool the machine. My criteria for a good handwriting analyzer is one that can recognize letters that are written on its side or upside-down. I bet there are some neural-nets out there that can do this. > So... to get the ball rolling... can anyone give me more scoop on the > BUMBLEBEE system... supposedly it contained roughly the same number of > neurons as a bumble-bee. I'd love to see it, it bet it flys! (no pun > intended... or was it :-) Sorry, but in terms of biological modelling, the best that has been done (that I know of) is a neural network that can simulate the movement of a round worm (Computer Simulation of the Motor-Neural System of a Simple Invertebrate, E. Niebur and P. Erdos). Sean Brunnock [[ For mailing list-only readers, this was in reponse to a message on USENET which otherwise contained little information and so was not included in the Digest. I believe the reference to the bumble-bee brain was from a recent MIT "long-range" study, which was quoted in the New York Times editorial section (7 Sept?). It said that in the near future (time undefiend) we should be able to build Neural Nets with the same number of elements as a bee's brain. The editorial quoted a military wag as saying, "There's more smarts in a bee than a cruise missile." -PM ]] ------------------------------ End of Neurons Digest *********************