From tgelder@phil.indiana.edu Fri Dec 1 02:11:35 1995 Received: from lucy.cs.wisc.edu by sea.cs.wisc.edu; Fri, 1 Dec 95 02:11:31 -0600; AA15317 Message-Id: <9512010811.AA12992@lucy.cs.wisc.edu> Received: from TELNET-1.SRV.CS.CMU.EDU by lucy.cs.wisc.edu; Fri, 1 Dec 95 02:11:29 -0600 Received: from TELNET-1.SRV.CS.CMU.EDU by telnet-1.srv.cs.CMU.EDU id ab28865; 29 Nov 95 12:00:34 EST Received: from GS151.SP.CS.CMU.EDU by TELNET-1.SRV.CS.CMU.EDU id ab28863; 29 Nov 95 11:40:10 EST Received: from GS151.SP.CS.CMU.EDU by GS151.SP.CS.CMU.EDU id aa26506; 29 Nov 95 11:40 EST Received: from CS.CMU.EDU by B.GP.CS.CMU.EDU id aa24436; 29 Nov 95 0:18:50 EST Received: from tarski.phil.indiana.edu by CS.CMU.EDU id aa05984; 29 Nov 95 0:17:58 EST Received: from [129.79.134.24] (tgelder.phil.indiana.edu) by phil.indiana.edu (4.1/9.7jsm) id AA16890; Wed, 29 Nov 95 00:17:54 EST Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Date: Wed, 29 Nov 1995 00:20:42 -0500 To: connectionists@cs.cmu.edu From: Tim van Gelder Subject: 'Mind as Motion' annct & web page Book announcement ::: Available now. `MIND AS MOTION: EXPLORATIONS IN THE DYNAMICS OF COGNITION' edited by Robert Port and Tim van Gelder Bradford Books/MIT Press. >From the dust jacket: `Mind as Motion is the first comprehensive presentation of the dynamical approach to cognition. It contains a representative sampling of original current research on topics such as perception, motor control, speech and language, decision making, and development. Included are chapters by pioneers of the approach, as well as others applying the tools of dynamics to a wide range of new problems. Throughout, particular attention is paid to the philosophical foundations of this radical new research program. Mind as Motion provides a conceptual and historical overview of the dynamical approach, a tutorial introduction to dynamics for cognitive scientists, and a glossary covering the most frequently used terms. Each chapter includes an introduction by the editors, outlining its main ideas and placing it in context, and a guide to further reading.' 668 pages, 139 illustrations ISBN 0-262-16150-8 $60.00 US For further information including the full text of the preface and a sample chapter introduction, see the web page at MIT Press: http://www-mitpress.mit.edu/mitp/recent-books/cog/mind-as-motion.html _________________________________________________ Chapter Titles and Authors 1.Tim van Gelder & Robert Port. Introduction: Its About Time: An Overview of the Dynamical Approach to Cognition. 2. Alec Norton. Dynamics: A Tutorial Introduction for Cognitive Scientists. 3. Esther Thelen. Time Scale Dynamics and the Development of an Embodied Cognition. 4. Jerome Busemeyer & James Townsend Dynamic Representation of Decision Making 5. Randall Beer Computational and Dynamical Languages for Autonomous Agents 6. Elliot Saltzman Dynamics and Coordinate Systems in Skilled Sensorimotor Activity 7. Catherine Browman & Louis Goldstein Dynamics and Articulatory Phonology 8. Jeffrey Elman Language as a Dynamical System 9. Jean Petitot Morphodynamics and Attractor Syntax 10. Jordan Pollack The Induction of Dynamical Recognizers 11. Paul van Geert Growth Dynamics in Development 12. Robert Port, Fred Cummins & Devin McAuley Naive Time, Temporal Patterns and Human Audition 13. Michael Turvey & Claudia Carello Some Dynamical Themes in Perception and Action 14. Geoffrey Bingham Dynamics and the Problem of Event Recognition 15. Stephen Grossberg Neural Dynamics of Motion Perception, Recognition Learning, and Spatial Attention 16. Mary Ann Metzger Multiprocess Models Applied to Cognitive and Behavioral Dynamics 17. Steven Reidbord & Dana Redington The Dynamics of Mind and Body During Clinical Interviews: Current Trends, Potential, and Future Directions 18. Marco Giunti Dynamical Models of Cognition 19. Glossary of Terminology in Dynamics _________________________________________________ At MIT Press, orders can be made by email at: mitpress-orders@mit.edu For general information re MIT Press, see: http://www-mitpress.mit.edu _________________________________________________ The editors welcome enquiries, discussion, critical feedback, etc. Robert Port Tim van Gelder Department of Linguistics Department of Philosophy Indiana University University of Melbourne Bloomington IN 47405 Parkville 3052 VIC USA AUSTRALIA port@indiana.edu tgelder@ariel.unimelb.edu.au From josh@faline.bellcore.com Fri Dec 1 02:11:37 1995 Received: from lucy.cs.wisc.edu by sea.cs.wisc.edu; Fri, 1 Dec 95 02:11:34 -0600; AA15321 Received: from TELNET-1.SRV.CS.CMU.EDU by lucy.cs.wisc.edu; Fri, 1 Dec 95 02:11:31 -0600 Received: from TELNET-1.SRV.CS.CMU.EDU by telnet-1.srv.cs.CMU.EDU id aa00751; 30 Nov 95 13:48:53 EST Received: from GS151.SP.CS.CMU.EDU by TELNET-1.SRV.CS.CMU.EDU id aa00749; 30 Nov 95 13:27:42 EST Received: from GS151.SP.CS.CMU.EDU by GS151.SP.CS.CMU.EDU id aa27157; 30 Nov 95 13:27 EST Received: from CS.CMU.EDU by B.GP.CS.CMU.EDU id aa04514; 29 Nov 95 13:20:35 EST Received: from faline.bellcore.com by CS.CMU.EDU id aa10766; 29 Nov 95 13:20:03 EST Received: (from josh@localhost) by faline.bellcore.com (8.6.9/8.6.10) id NAA15771 for connectionists@cs.cmu.edu; Wed, 29 Nov 1995 13:19:47 -0500 Date: Wed, 29 Nov 1995 13:19:47 -0500 From: Joshua Alspector Message-Id: <199511291819.NAA15771@faline.bellcore.com> To: connectionists@cs.cmu.edu Subject: Research associate in neuromorphic electronics RESEARCH ASSOCIATE IN NEUROMORPHIC ELECTRONICS There is an anticipated position in the electrical and computer engineering department at the University of Colorado at Colorado Springs for a postdoctoral research associate in the area of neural learning microchips. The successful candidate will have experience in analog and digital VLSI design and test and be comfortable working at the system level in a UNIX/C/C++ environment. The project will involve applying an existing VME-based neural network learning system to several demanding problems in signal processing. These include adaptive non-linear equalization of underwater acoustic communication channels and magnetic recording channels. It is likely also to involve integrating the learning electronics with micro-machined sonic transducers directly on silicon. Please send a curriculum vita, names and addresses of at least three referees, and and copies of some representative publications to: Prof. Joshua Alspector Univ. of Colorado at Col. Springs Dept. of Elec. & Comp. Eng. P.O. Box 7150 Colorado Springs, CO 80933-7150 From ajit@austin.ibm.com Fri Dec 1 02:11:39 1995 Received: from lucy.cs.wisc.edu by sea.cs.wisc.edu; Fri, 1 Dec 95 02:11:35 -0600; AA15325 Received: from TELNET-1.SRV.CS.CMU.EDU by lucy.cs.wisc.edu; Fri, 1 Dec 95 02:11:33 -0600 Received: from TELNET-1.SRV.CS.CMU.EDU by telnet-1.srv.cs.CMU.EDU id ab00773; 30 Nov 95 14:00:01 EST Received: from GS151.SP.CS.CMU.EDU by TELNET-1.SRV.CS.CMU.EDU id aa00755; 30 Nov 95 13:30:04 EST Received: from GS151.SP.CS.CMU.EDU by GS151.SP.CS.CMU.EDU id aa27165; 30 Nov 95 13:29 EST Received: from CS.CMU.EDU by B.GP.CS.CMU.EDU id aa05916; 29 Nov 95 14:47:30 EST Received: from netmail1.austin.ibm.com by CS.CMU.EDU id aa11877; 29 Nov 95 14:46:31 EST Received: from ding.austin.ibm.com (ding.austin.ibm.com [129.35.221.140]) by netmail1.austin.ibm.com (8.6.12/8.6.11) with SMTP id NAA52534 for ; Wed, 29 Nov 1995 13:46:04 -0600 Received: by ding.austin.ibm.com (AIX 3.2/UCB 5.64/4.03-client-2.6) for Connectionists@CS.CMU.EDU at austin.ibm.com; id AA12676; Wed, 29 Nov 1995 13:46:04 -0600 Date: Wed, 29 Nov 1995 13:46:04 -0600 From: Dingankar Message-Id: <9511291946.AA12676@ding.austin.ibm.com> To: Connectionists@CS.cmu.edu Subject: "Network Approximation of Dynamical Systems" - Neuroprose paper Reply-To: ajit@austin.ibm.com Organization: IBM, Austin Work: (512) 838-6850 Fax: (512) 838-5882 Ftp-Host: archive.cis.ohio-state.edu Ftp-Filename: /pub/neuroprose/dingankar.tensor-products.ps.Z **DO NOT FORWARD TO OTHER GROUPS** Sorry, no hardcopies available. 6 pages. Greetings! The following invited paper will be presented at NOLTA'95 next month. The compressed PostScript file is available in the Neuroprose archive; the details (URL, bibtex entry and abstract) follow. Thanks, Ajit ------------------------------------------------------------------------------ URL: ftp://archive.cis.ohio-state.edu/pub/neuroprose/dingankar.tensor-products.ps.Z BiBTeX entry: @INPROCEEDINGS{atd:nolta-95, AUTHOR ="Dingankar, Ajit T. and Sandberg, Irwin W.", TITLE ="{Network Approximation of Dynamical Systems}", BOOKTITLE ="Proceedings of the International Symposium on Nonlinear Theory and its Applications (NOLTA'95)", YEAR ="1995", EDITOR ="", PAGES ="", ORGANIZATION ="", PUBLISHER ="", ADDRESS ="Las Vegas, Nevada", MONTH ="December 10--14" } Network Approximation of Dynamical Systems ------------------------------------------ ABSTRACT We consider the problem of approximating any member of a large class of input-output operators of time-varying nonlinear dynamical systems. We introduce a family of ``tensor product" dynamical neural networks, and show that a certain continuity condition is necessary and sufficient for the existence of arbitrarily good approximations using this family. ------------------------------------------------------------------------------ Ajit T. Dingankar | ajit@austin.ibm.com IBM Corporation, Internal Zip 4359 | Work: (512) 838-6850 11400 Burnet Road, Austin, TX 78758 | Fax : (512) 838-5882 From krose@fwi.uva.nl Fri Dec 1 23:22:48 1995 Received: from lucy.cs.wisc.edu by sea.cs.wisc.edu; Fri, 1 Dec 95 23:22:44 -0600; AA02735 Received: from TELNET-1.SRV.CS.CMU.EDU by lucy.cs.wisc.edu; Fri, 1 Dec 95 23:22:42 -0600 Received: from TELNET-1.SRV.CS.CMU.EDU by telnet-1.srv.cs.CMU.EDU id aa00773; 30 Nov 95 13:58:54 EST Received: from GS151.SP.CS.CMU.EDU by TELNET-1.SRV.CS.CMU.EDU id aa00753; 30 Nov 95 13:29:16 EST Received: from GS151.SP.CS.CMU.EDU by GS151.SP.CS.CMU.EDU id aa27161; 30 Nov 95 13:28 EST Received: from CS.CMU.EDU by B.GP.CS.CMU.EDU id aa21440; 30 Nov 95 8:27:34 EST Received: from mail.fwi.uva.nl by CS.CMU.EDU id aa18506; 30 Nov 95 8:26:16 EST Received: from carol.fwi.uva.nl by mail.fwi.uva.nl with ESMTP (sendmail 8.6.12/config 5.15). id OAA14649; Thu, 30 Nov 1995 14:25:12 +0100 Return-Path: Received: from localhost by carol.fwi.uva.nl (sendmail 8.6.12/config 5.15). id OAA08623; Thu, 30 Nov 1995 14:25:11 +0100 Message-Id: <199511301325.OAA08623@carol.fwi.uva.nl> From: Ben Krose X-Organisation: Faculty of Mathematics, Computer Science, Physics & Astronomy University of Amsterdam Kruislaan 403 NL-1098 SJ Amsterdam The Netherlands X-Phone: +31 20 525 7463 X-Fax: +31 20 525 7490 Subject: paper announcement To: Connectionists@CS.cmu.edu Date: Thu, 30 Nov 1995 14:25:10 +0100 (MET) X-Mailer: ELM [version 2.4 PL23] Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Content-Length: 2014 The following paper, published at the International Conference on Artificial Neural Networks 1995 (ICANN*95), Springer Verlag, is available from: http://www.fwi.uva.nl/research/neuro/publications/publications.html or ftp://ftp.fwi.uva.nl/pub/computer-systems/aut-sys/reports/VysGroKro95.ps.gz The paper is 6 pages long, in compressed form 52kB. Sorry, no hardcopies available. =========================================================================== ``Orthogonal incremental learning of a feedforward network'' V. Vysniauskas, F.C.A. Groen and B.J.A. Kr\"ose (1995) Proceedings of the 1995 International Conference on Neural Networks, Edited by Fogelman-Soulie and Gallinari, pp. 311-317. Abstract Orthogonal incremental learning (OIL) is a new approach of incremental training for a feedforward network with a single hidden layer. OIL is based on the idea to describe the output weights (but not the hidden nodes) as a set of orthogonal basis functions. Hidden nodes are treated just as the orthogonal representation of the network in the output weights domain. We showed that the network training can be performed incrementally, one node at time, and there is no need to use an additional constraint to support a consistent optimization among the hidden nodes. An advantage of OIL over existing algorithms is extremely fast learning. This approach can be also easily extended to build-up incrementally an arbitrary function as a linear composition of adjustable functions which are not necessarily orthogonal. We tested this approach on a standard "two-spirals" benchmark problem to build incrementally a feedforward network with a single layer of Gaussian units. ============================================================================ We welcome your comments, Ben Kr\"ose Faculty of Mathematics and Computer Science University of Amsterdam Kruislaan 403 1098 SJ Amsterdam Netherlands tel: +31 20 525 7520/7463 krose@fwi.uva.nl http://www.fwi.uva.nl/research/neuro/ From dhw@santafe.edu Sat Dec 2 10:04:23 1995 Received: from lucy.cs.wisc.edu by sea.cs.wisc.edu; Sat, 2 Dec 95 10:04:19 -0600; AA08812 Received: from TELNET-1.SRV.CS.CMU.EDU by lucy.cs.wisc.edu; Sat, 2 Dec 95 10:04:16 -0600 Received: from TELNET-1.SRV.CS.CMU.EDU by telnet-1.srv.cs.CMU.EDU id aa02535; 1 Dec 95 12:22:59 EST Received: from GS151.SP.CS.CMU.EDU by TELNET-1.SRV.CS.CMU.EDU id aa02533; 1 Dec 95 12:03:14 EST Received: from GS151.SP.CS.CMU.EDU by GS151.SP.CS.CMU.EDU id aa27944; 1 Dec 95 12:02 EST Received: from CS.CMU.EDU by B.GP.CS.CMU.EDU id av13873; 1 Dec 95 11:27:53 EST Received: from sfi.santafe.edu by CS.CMU.EDU id aa29741; 1 Dec 95 11:20:53 EST Received: from santafe (santafe.santafe.edu) by sfi.santafe.edu (4.1/SMI-4.1) id AA27395; Fri, 1 Dec 95 09:18:19 MST Date: Fri, 1 Dec 95 09:18:19 MST From: David Wolpert Message-Id: <9512011618.AA27395@sfi.santafe.edu> To: Connectionists@CS.cmu.edu Subject: Correcting misunderstandings about NFL This posting is to correct some misunderstandings that were recently posted concerning the NFL theorems. I also draw attention to some of the incorrect interpretations commonly ascribed to certain COLT results. *** Joerg Lemm writes: >>> 1.) If there is no relation between the function values on the test and training set (i.e. P(f(x_j)=y|Data) equal to the unconditional P(f(x_j)=y) ), then, having only training examples y_i = f(x_i) (=data) from a given function, it is clear that I cannot learn anything about values of the function at different arguments, (i.e. for f(x_j), with x_j not equal to any x_i = nonoverlapping test set). >>> Well put. Now here's the tough question: Vapnik *proves* that it is unlikely (for large enough training sets and small enough VC dimension generalizers) for error on the training set and full "generalization error" to be grealy different. Regardless of the target. Using this, Baum and Haussler even wrote a paper "What size net gives valid generalization?" in which no assumptions whatsoever are made about the target, and yet the authors are able to provide a response the question of their title. HOW IS THAT POSSIBLE GIVEN WHAT YOU JUST WROTE???? NFL is "obvious". And so are VC bounds on generalization error (well, maybe not "obvious"). And so is the PAC "proof" of Occam's razor. And yet the latter two bound generalization error (for those cases where training set error is small enough) without making any assumptions about the target. What gives? The answer: The math of those works is correct. But far more care must be exercised in the interpretation of that math than you will find in those works. The care involves paying attention to what goes on the right-hand side of the conditioning bars in one's probabilities, and the implications of what goes there. Unfortunately, such conditioning bar are completely absent in those works... (In fact, the sum-total of the difference between Bayesian and COLT approaches to supervised batch learning lies in what's on the right-hand side of those bars, but that's another story. See [2].) As an example, it is widely realized that VC bounds suffer from being worst-case. However there is another hugely important caveat to those bounds. The community as a whole simply is not aware of that caveat, because the caveat concerns what goes on the right-hand side of the conditioning bar, and this is NEVER made explicit. This caveat is the fact that VC bounds do NOT concern Pr(IID generalization error | observed error on the training set, training set size, VC dimension of the generalizer). But you wouldn't know that to read the claims made on behalf of those bounds ... To give one simple example of the ramifications of this: Let's say you have a favorite low-VC generalizer. And in the course of your career you parse though learning problems, either explicitly or (far more commonly) without even thinking about it. When you come across one with a large training set on which your generalizer has small generalization error, you want to invoke Vapnik to say you have assuraces about full generalization error. Well, sorry. You don't and you can't. You simply can't escape Bayes by using confidence intervals. Confidence intervals in general (not just in VC work) have the annoying property that as soon as you try to use them, very often you contradict the underlying statistical assumptions behind them. Details are in [1] and in the discussion of "We-Learn-It Inc." in [2]. >>> 2.) We are considering two of those (influence) relations P(f(x_j)=y|Data): one, named A, for the true nature (=target) and one, named B, for our model under study (=generalizer). Let P(A and B) be the joint probability distribution for the influence relations for target and generalizer. 3.) Of course, we do not know P(A and B), but in good old Bayesian tradition, we can construct a (hyper-)prior P(C) over the family of probability distributions of the joint distributions C = P(A and B). 4.) NFL now uses the very special prior assumption P(A and B) = P(A)P(B) >>> If I understand you correctly, I would have to disagree. NFL also holds with your P(C) being any prior assumption - more formally, averaging over all priors, you get NFL. So the set of priors for which your favorite algorithm does *worse than random* is just as large as the set for which it does better. (In this sense, the uniform prior is a typical prior, not a pathological one, out on the edge of the space. It is certainly not a "very special prior".) In fact, that's one of the major points of NFL - it's not to see what life would be like if this or that were uniform, but to use such uniformity as a mathematical tool, to get a handle on the underlying geometry of inference, the size of the various spaces (e.g., the size of the space of priors for which you lose to random), etc. The math *starts* with NFL, and then goes on to many other things (see [1]). It's only the beginning chapter of the text book. >>> I say that it is rational to believe (and David does so too, I think) that in real life cross-validation works better in more cases than anti-cross-validation. >>> Oh, most definitely. There are several issues here: 1) what gives with all the "prior-free" general proofs of COLT, given NFL, 2) purely theoretical issues (e.g., as mentioned before, characterizing the relationship between target and generalizers needed for xval. to beat anti-xval.) and 3) perhaps most provocatively of all, seeing if NFL (and the associated mathematical structure) can help you generalize in the real world (e.g., with head-to-head minimax distinctions between generalizers). *** Finally, Eric Baum weighs in: >>> Barak Pearlmutter remarked that saying We have *no* a priori reason to believe that targets with "low Kolmogorov complexity" (or anything else) are/not likely to occur in the real world. (which I gather was a quote from David Wolpert?) is akin to saying we have no a priori reason to believe there is non-random structure in the world, which is not true, since we make great predictions about the world. >>> Well, let's get a bit formal here. Take all the problems we've ever tried to make "great predictions" on. Let's even say that these problems were randomly chosen from those in the real world (i.e., no selection effects of people simply not reporting when their predictions were not so great). And let's for simplicity say that all the predictions were generated by the same generalizer - the algorithm in the brain of Eric Baum will do as a straw man. Okay. Now take all those problems together and view them as one huge training set. Better still, add in all the problems that Eric's anscestors addressed, so that the success of his DNA is also taken into account. That's still one training set. It's a huge one, but it's tiny in comparison to the full spaces it lives in. Saying we (Eric) makes "great predictions" simply means that the xvalidation error of our generalizer (Eric) on that training set is small. (You train on part of the data, and predict on the rest.) Formally (!!!!!), this gives no assuraces whatsoever about any behavior off-training-set. As I've stated before, without assumptions, you cannot conclude that low xvalidation error leads to low off-training set generalization error. And of course, each passing second, each new scene you view, is "off-training-set". The fallacy in Eric's claim was noted all the way back by Hume. Success at inductive inference cannot formally establish the utility of using inductive inference. To claim that it can you have to invoke inductive inference, and that, as any second grader can tell you, is circular reasoning. Practically speaking of course, none of this is a concern in the real world. We are all (me included) quite willing to conclude there is structure in the real world. But as was noted above, what we do in practice is not the issue. The issue is one of theory. *** It's very similar to high-energy physics. There are a bunch of physical constants that, if only slightly varied, would (seem to) make life impossible. Why do they have the values they have? Some invoke the anthropic principle to answer this - we wouldn't be around if they had other values. QED. But many find this a bit of a cop-out, and search for something more fundamental. After all, you could have stopped the progress of physics at any point in the past if you had simply gotten everyone to buy into the anthropic principle at that point in time. Similarly with inductive inference. You could just cop out and say "anthropic principle" - if inference were not possible, we wouldn't be having this debate. But that's hardly a satisfying answer. *** Eric goes on: >>> Consider the problem of learning to predict the pressure of a gas from its temperature. Wolpert's theorem, and his faith in our lack of prior about the world, predict, that any learning algorithm whatever is as likely to be good as any other. This is not correct. >>> To give two examples from just the past month, I'm sure MCI and Coca-Cola would be astonished to know that the algorithms they're so pleased with were designed for them by someone having "faith in our lack of prior about the world". Less glibly, let me address this claim about my "faith" with two quotes from the NFL for supervised learning paper. The first is in the introduction, and the second in a section entitled "On uniform averaging". So neither is exactly hidden ... 1) "It cannot be emphasized enough that no claim is being made .. that all algorithms are equivalent in the real world." 2) "The uniform sums over targets ... weren't chosen because there is strong reason to believe that all targets are equally likely to arise in practice. Indeed, in many respects it is absurd to ascribe such a uniformity over possible targets to the real world. Rather the uniform sums were chosen because such sums are a useful theoretical tool with which to analyze supervised learning." Finally, given that I'm mixing it up with Eric on NFL, I can't help but quote the following from his "What size net gives valid generalization" paper: "We have given bounds (independent of the target) on the training set size vs. neural net size need such that valid generalization can be expected." (Parenthetical comment added - and true.) Nowhere in the paper is there any discussion whatsoever of the apparent contradiction between this statement and NFL-type concerns. Indeed, as mentioned above, with only the conditioning-bar-free mathematics in Eric's paper, there is no way to resolve the contradiction. In this particular sense, that paper is extremely misleading. (See discussion above on misinterpretations of Vapnik's results.) >>>> Creatures evolving in this "play world" would exploit this structure and understand their world in terms of it. There are other things they would find hard to predict. In fact, it may be mathematically valid to say that one could mathematically construct equally many functions on which these creatures would fail to make good predictions. But so what? So would their competition. This is not relevant to looking for one's key, which is best done under the lamppost, where one has a hope of finding it. In fact, it doesn't seem that the play world creatures would care about all these other functions at all. >>> I'm not sure I quite follow this. In particular, the comment about the "competition" seems to be wrong. Let me just carry further Eric's metaphor though, and point out though that it makes a hell of a lot more sense to pull out a flashlight and explore into the surrounding territory for your key than it does to spend all your time with your head down, banging into the lamppost. And NFL is such a flashlight. David Wolpert [1] The current versions of the NFL for supervised learning papers, nfl.ps.1.Z and nfl.ps.2.Z, at ftp.santafe.edu, in pub/dhw_ftp. [2] "The Relationship between PAC, the Statistical Physics Framework, the Bayesian Framework, and the VC Framework", in *The Mathematics of Generalization*, D. Wolpert Ed., Addison-Wesley, 1995. From marco@McCulloch.Ing.UniFI.IT Sat Dec 2 19:56:10 1995 Received: from lucy.cs.wisc.edu by sea.cs.wisc.edu; Sat, 2 Dec 95 19:56:07 -0600; AA13356 Received: from TELNET-1.SRV.CS.CMU.EDU by lucy.cs.wisc.edu; Sat, 2 Dec 95 19:56:05 -0600 Received: from TELNET-1.SRV.CS.CMU.EDU by telnet-1.srv.cs.CMU.EDU id aa04736; 2 Dec 95 18:49:42 EST Received: from GS151.SP.CS.CMU.EDU by TELNET-1.SRV.CS.CMU.EDU id aa04734; 2 Dec 95 18:32:43 EST Received: from GS151.SP.CS.CMU.EDU by GS151.SP.CS.CMU.EDU id aa29589; 2 Dec 95 18:31 EST Received: from RI.CMU.EDU by B.GP.CS.CMU.EDU id aa18886; 1 Dec 95 16:14:08 EST Received: from cesit1.unifi.it by RI.CMU.EDU id aa25781; 1 Dec 95 16:12:54 EST Received: from McCulloch.Ing.UniFI.IT by CESIT1.UNIFI.IT (PMDF V5.0-4 #3688) id <01HYAWOSUHY8000Q3Z@CESIT1.UNIFI.IT> for Connectionists@CS.CMU.EDU; Fri, 01 Dec 1995 18:25:02 +0100 (MET) Received: by McCulloch.Ing.UniFI.IT (5.x/SMI-SVR4) id AA09634; Fri, 01 Dec 1995 18:21:43 +0100 Date: Fri, 01 Dec 1995 18:21:43 +0100 From: Marco Gori Subject: Italian Neural Network Society To: Connectionists@CS.cmu.edu Message-Id: <9512011721.AA09634@McCulloch.Ing.UniFI.IT> Organization: DSI - University of Florence (Italy) Content-Transfer-Encoding: 7BIT X-Sun-Charset: US-ASCII ============================================================== This is to announce a new web page describing the aims and the activities of the Italian Neural Network Society. The page is hosted at the DSI Web server of the Dipartimento di Sistemi e Informatica, Universita' di Firenze) at the following address: http://www-dsi.ing.unifi.it/neural/siren -- marco gori. ===============================================================