From dave@twinearth.wustl.edu Sat Apr 1 00:27:05 1995 Received: from lucy.cs.wisc.edu by sea.cs.wisc.edu; Sat, 1 Apr 95 00:27:00 -0600; AA06326 Received: from TELNET-1.SRV.CS.CMU.EDU by lucy.cs.wisc.edu; Sat, 1 Apr 95 00:26:57 -0600 Received: from TELNET-1.SRV.CS.CMU.EDU by telnet-1.srv.cs.CMU.EDU id aa22408; 31 Mar 95 15:34:04 EST Received: from DST.BOLTZ.CS.CMU.EDU by TELNET-1.SRV.CS.CMU.EDU id aa22406; 31 Mar 95 15:13:58 EST Received: from DST.BOLTZ.CS.CMU.EDU by DST.BOLTZ.CS.CMU.EDU id aa21045; 31 Mar 95 15:13:46 EST Received: from CS.CMU.EDU by B.GP.CS.CMU.EDU id aa28991; 31 Mar 95 3:53:03 EST Received: from twinearth.wustl.edu by CS.CMU.EDU id aa23297; 31 Mar 95 3:52:13 EST Received: by twinearth.wustl.edu (4.1/SMI-SVR4) id AA11316; Fri, 31 Mar 95 02:48:35 CST Date: Fri, 31 Mar 95 02:48:35 CST From: David Chalmers Message-Id: <9503310848.AA11316@twinearth.wustl.edu> To: connectionists@cs.cmu.edu, philos-l@liverpool.ac.uk, philosop@yorkvm1.bitnet, spp@umiacs.umd.edu Subject: Penrose symposium Over the coming weeks, there will be a symposium on Roger Penrose's recent book SHADOWS OF THE MIND in the electronic journal PSYCHE. There will be ten review articles discussing the issues raised by the book from a variety of perspectives, and Penrose will reply. Authors of the review articles are: Bernard Baars: Psychology, The Wright Institute, Berkeley David Chalmers: Philosophy, Washington University Solomon Feferman: Mathematics, Stanford University Stanley Klein: Vision Sciences, University of California at Berkeley Aaron Klug: Molecular Biology, Cambridge University Tim Maudlin: Philosophy, Rutgers University John McCarthy: Computer Science, Stanford University Daryl McCullough: Computer Science, Odyssey Research Associates Drew McDermott: Computer Science, Yale University Hans Moravec: Robotics, Carnegie Mellon University The articles will appear at a rate of one or two per week, and will be e-mailed to subscribers of the mailing list PSYCHE-L. To subscribe to this mailing list, send e-mail to listserv@iris.rfmh.org, with a single line "SUBSCRIBE PSYCHE-L ". The articles will also be made available on the worldwide web, at http://hcrl.open.ac.uk/psyche.html (this is the PSYCHE home page). Discussion is encouraged on the associated discussion list PSYCHE-D -- to subscribe, "SUBSCRIBE PSYCHE-D" at the address above. David Chalmers. From marius@crab.psy.cmu.edu Sat Apr 1 13:11:57 1995 Received: from lucy.cs.wisc.edu by sea.cs.wisc.edu; Sat, 1 Apr 95 13:10:52 -0600; AA22609 Received: from TELNET-1.SRV.CS.CMU.EDU by lucy.cs.wisc.edu; Sat, 1 Apr 95 13:09:55 -0600 Received: from TELNET-1.SRV.CS.CMU.EDU by telnet-1.srv.cs.CMU.EDU id aa22786; 31 Mar 95 20:50:36 EST Received: from DST.BOLTZ.CS.CMU.EDU by TELNET-1.SRV.CS.CMU.EDU id aa22783; 31 Mar 95 20:42:31 EST Received: from DST.BOLTZ.CS.CMU.EDU by DST.BOLTZ.CS.CMU.EDU id aa21346; 31 Mar 95 20:41:46 EST Received: from EDRC.CMU.EDU by B.GP.CS.CMU.EDU id aa12484; 31 Mar 95 20:26:26 EST Received: from ANDREW.CMU.EDU by EDRC.CMU.EDU id aa11759; 31 Mar 95 20:25:07 EST Received: from crab.psy.cmu.edu.psy.cmu.edu (CRAB.PSY.CMU.EDU [128.2.248.139]) by andrew.cmu.edu (8.6.9/8.6.9) with SMTP id UAA13479 for ; Fri, 31 Mar 1995 20:25:01 -0500 Received: by crab.psy.cmu.edu.psy.cmu.edu (4.1/SMI-4.1) id AA12302; Fri, 31 Mar 95 20:25:00 EST Date: Fri, 31 Mar 95 20:25:00 EST From: Marius Usher Message-Id: <9504010125.AA12302@crab.psy.cmu.edu.psy.cmu.edu> To: Connectionists@cs.cmu.edu Subject: discussion on variability and neural codes Perhaps the most crucial question in the study of cortical function is whether the brain uses a mean rate code or a temporal code. Recently a number of models have been proposed in order to account for the variability of spike trains (discussed by Softky and Koch, 1993). As it seems, each of these models can account for variability, despite their very different assumptions and implications regarding the "neural code". We are writing this note in order to highlight the specific predictions in which these models differ, hoping in particular to direct the attention of experimentalists to the "missing " data required to disambiguate between these theoretical models and their implications about the neural code. In the following, we briefly outline three type of models proposed to account for variability, discuss their different predictions and implications and finally suggest experimental tests to distinguish between them. 1. Random-walks. Shadlen and Newsome (1994) proposed that the variability problem had in fact been solved by the random walk model of Gerstein and Mandelbrot (1964) thirty years ago. This model relies on well-balanced excitation and inhibition to a single cell, so that its membrane potential performs a random walk without bias and the cell fires when the random walk reaches threshold, after which it is reset. They have shown that the model is consistent with neurophysiological estimated parameters, that it produces high variability, and they claimed that it implies a neural code based on rate modulations. 2. Coincidence detection. Bell et al. (1995; see also message to "Connectionists" last week) showed that the coincidence detection mechanism originally postulated by Softky and Koch (1993) to explain variability can be obtained in neurons using biophysically realistic parameters. Their model also relies on well-balanced excitation and inhibition to the single cell, in addition to fast membrane time constants and very weak after-depolarization. At even weaker levels of afterdepolarization, cells show a bursting tendency which further increases variability. Another coincidence detection model on the basis of active sodium dendritic currents has been proposed by Softky (1994). 3. Internally generated rate fluctuations with partial synchronization. We have proposed a model (Usher et al. 1994;1995) in which the neural system generates its own rate fluctuations (even under stationary stimulation), in combination with partially synchronized activity (at 5-10 msec). These two factors interact: partial synchronization de-stabilizes the system, enhancing rate fluctuations. The model relies on local center-surround (excitation/inhibition) circuitry resulting in moving clusters of activity in the modeled cortical layer. The model can work with well-balanced excitation and inhibition, but does not require it in order to produce high variability spike trains. Other models which can also generate rate fluctuations (on the basis of spontaneous symmetry breaking in orientation-preference space) have been proposed recently by Hansel and Sompolinsky (1995) and by Somers et al. (1995), in order to explain the enhancement of orientation tuning in visual cortex. Predictions: Variability, as measured by the coefficient of variance CV (standard deviation/mean) is not sufficient to distinguish between these models. Higher order statistics (form of ISI distributions: exponential vs. power-laws, serial correlations between intervals, auto-correlation, etc) are required: 1. The balanced random walk (with leakage) and the coincidence detection model, predict that ISI distributions have exponential tails and that there are no serial correlations between intervals (unless in the bursting regime of the coincidence detection model). Both models should be indistinguishable from a Poisson process with a refractory period (correlation functions should be flat at long times relative to the refractory period). 2. The rate fluctuation model produces (in some parameter regime) interval distributions with power laws (whose exponents vary depending on the network parameters), non-zero serial correlation between intervals, and correlation function with a dual temporal structure (castle-on-hill). This model also predicts oscillatory local field potentials even when single cells fire irregularly. Data and Implications Data regarding the form of interval distributions is unfortunately sparse. While some reports show exponentially distributed intervals, other studies found power-law distributions (Lowen and Teich 1992,Gruneis et al. 1990, Wise 1981). Another feature of data in support of the rate fluctuation models is provided by typical correlation functions with dual temporal structure (castle-on-hill) (Nelson et al. 1992), that can not be explained by the coincidence detection or random walk models. On the other hand, cells in visual cortex in vivo show weak afterhyperpolarization supporting the claim of Bell et al. (1995), however intracellular voltage traces also show slow variations in resting potential, consistent with rate fluctuations models. It is not yet clear how these factors vary with brain system and with type of stimulation. Systematic experimental studies are thus needed to decide on the principles of information processing used by various brain systems. It should be critical to measure whether cortical cells in vivo, are under well-balanced excitation/inhibition input. Studies by Douglas et al. (1988) and Nelson et al. (1994) seem to indicate the opposite, but a definitive answer would require a study that specifically blocks all GABAergic input to a single cell in vivo without affecting the spike repolarization currents. The coincidence detection principle supports a temporal code: the apparent irregularity of spike trains hides information in the exact time intervals between spikes, like a Morse signal. This may require a system that is intrinsically deterministic, which may be problematic in light of internal fluctuations of nerve cells (e.g., failure of neurotransmitter release). On the other hand, the random walk and rate fluctuation models, support a rate code; a difference between the two, is that due to its internal instability, the rate fluctuation model is able to amplify small fluctuations in the input. In Usher et al. (1995) we have shown that the in the range of parameters where power-laws are obtained, the system achieves a trade-of between two conflicting demands: fast amplification of input fluctuations and memory. Synchronous neural activity at a 5-10 msec time scale (unlike at submsec scale) is robust, and may play a role in the processes underlying attention and binding (Niebur and Koch 1994) (but no Morse codes are required for this). While the extensive body of data showing selective sensory responses and sustained activity in delayed tasks (Fuster and Jervey 1980, Myashita and Chang 1988) indicate a neural rate code, some support for coincidence detection has been provided by the "coincident sequences" measured by Abeles et al. (1993). However, neither of the models discussed here, can predict such "delayed coincidences". (For this, a synfire type model is required, but no ISI distributions are yet available for this type of model). Proponents of the coincidence detection principle may need to find an explanation for the wealth of evidence showing integration in the perceptual system. For example, the Bloch law (Loftus and Ruthruff 1993) shows that, for stimuli of duration shorter than 100 msec, perceptual discrimination depends only on the INTEGRAL of the stimulus (a high contrast 10 msec stimuli, produces exactly the same affect on perception as a 20 msec stimuli of half contrast). Marius Usher Martin Stemmler Dept. of Psychology, Computational and Neural Systems Carnegie Mellon University California Institute of Technology Our References: Usher M, Stemmler M., Koch C., and Olami Z (1994). "Network amplification of local fluctuations may cause high spike rate variability and fractal neuronal firing patterns". Neur. Comp. 6, 795-836. Usher M., Stemmler M, and Christof Koch (1994). "Oscillatory field potentials in the presence of irregular discharge patterns". Proceedings of the Computational and Neural Systems Conference, Monterey, July 1994. Usher M., Stemmler M. and Olami Z. (1995) "Dynamic pattern formation leads to $1/f$ noise in neural populations", Phys.Rev.Lett., 74, 326-329. Our papers can be also accessed by ftp from: FTP-host: klab.caltech.edu (anonymous, binary) FTP-filename: /pub/usher/NC.ps.gz /pub/usher/CNS94.ps.gz /pub/usher/prl.ps.gz (files need to be uncompressed with gunzip, in unix) (since some people may have difficulties printing the files due to the figures, versions of the papers without figures are placed in the same directory) Other References: Abeles M., Bergman H, Margalit E and Vaadia E. (1993). "Spatiotemporal firing patterns in the frontal cortex of behaving monkeys". J.Neurophys.,70,1629-1638 Bair W, Koch C, Newsome W and Britten K (1995). "Power spectrum analysis of bursting cells in area MT in behaving monkey". J. of Neurosc.,14, 2870-2892 Bell AJ, Mainen ZF, Tsodyks M and Sejnowski T (1995). "Balancing of conductances may explain irregular cortical spiking". Technical Report no. INC-9502 UCSD ******FTP: (SITE: ftp.salk.edu, FILE: pub/tony/bell.noisy.ps.Z)****** Douglas RJ. Martin KA and Whitteridge D (1988). "Selective responses of visual cortical cells do not depend on shunting inhibition", Nature, 332, 642-644. Fuster JM and Jervey JP (1980). "Inferotemporal neurons distinguish and retain behaviorally relevant features of visual stimuli", Science, 212, 952-955. Gerstein G and Mandelbrot (1964). "Random walk models for spike activity in single neurons", Biophys. J. 4, 41-65. Gruneis F, Nakao M and Yamakato M (1990). "Counting statistics of 1/f fluctuations a in neuronal spike trains". Biol. Cybern. 62, 407-413. Hansel D and Sompolinsky H (1995). "Chaos and synchrony in a model of a hypercolumn in visual cortex", Technical report, Racah Institute and C.P.T. Ecole Polytechnique. Loftus G. and Ruthruff E (1994). "A theory of visual information acquisition and visual memory with special application to intensity duration trade-offs. J. of Exp. Psych. HPP. Lowen SB and Teich MC (1992) "Auditory-nerve action potentials from a non-renewal point process over short as well as long time scales", J. Acoustical Soc. of Am., 92, 803-806. Miyashita Y. and Chang HS (1988). "Neural correlate of pictorial short term memory", Nature, 331, 68-70. Nelson JI, Salin PA, Munk MH, Arzi M (1992) "A cross correlation study in areas 17 and 18 of the cat", Visual Neurosc., 9, 21-38. Nelson S, Toth L., Sheth B and Sur M (1994). "Orientation selectivity of cortical neurons during intracellular blocking of inhibition", Science,265,744-777 Niebur E and Koch C (1994). "A model for the neuronal implementation of selective visual attention based on temporal correlations among neurons". J. Comp. Neurosc. 1, 141-158. Shadlen MN and Newsome WT (1994). "Noise, neural codes and cortical organization", Current Opinion in Neurobiol., 4, 569-579. Softky WR and Koch C (1993). "The highly irregular firing of cortical cells is inconsistent with temporal integration of EPSPs. J. Neurosc., 13, 334-350. Softky WR (1994). "Submillisecond coincidence detection in active dendritic trees". Neurosc. 58, 13-41. Somers D, Nelson S and Sur M. (1995). "An emergent model of orientation selectivity in cat visual cortical simple cells", J. of Neurosc., in press. Wise ME (1981). "Spike distributions for neurons and random walks with drift to a fluctuating threshold". Statistical Distributions in scientific work, vol. 6, eds. C Taillie et al. (D. Reidel, Dordrecht-Holland). From bishopc@helios.aston.ac.uk Sat Apr 1 13:12:05 1995 Received: from lucy.cs.wisc.edu by sea.cs.wisc.edu; Sat, 1 Apr 95 13:11:12 -0600; AA22611 Received: from TELNET-1.SRV.CS.CMU.EDU by lucy.cs.wisc.edu; Sat, 1 Apr 95 13:07:20 -0600 Received: from TELNET-1.SRV.CS.CMU.EDU by telnet-1.srv.cs.CMU.EDU id ac22415; 31 Mar 95 15:47:01 EST Received: from DST.BOLTZ.CS.CMU.EDU by TELNET-1.SRV.CS.CMU.EDU id ac22410; 31 Mar 95 15:15:08 EST Received: from DST.BOLTZ.CS.CMU.EDU by DST.BOLTZ.CS.CMU.EDU id aa21061; 31 Mar 95 15:14:38 EST Received: from EDRC.CMU.EDU by B.GP.CS.CMU.EDU id aa07878; 31 Mar 95 15:08:22 EST Received: from [134.151.79.46] by EDRC.CMU.EDU id aa10529; 31 Mar 95 15:05:17 EST Received: from email.aston.ac.uk by hermes.aston.ac.uk with SMTP (PP); Fri, 31 Mar 1995 20:51:56 +0100 Received: from synapse by email.aston.ac.uk with SMTP (PP) id <16175-0@email.aston.ac.uk>; Fri, 31 Mar 1995 20:46:40 +0100 From: bishopc Date: Fri, 31 Mar 1995 20:53:24 +0000 Message-Id: <4008.9503311953@sun.aston.ac.uk> To: Connectionists@cs.cmu.edu Subject: Lectureships in Neural Computing Cc: c.m.bishop@aston.ac.uk X-Sun-Charset: US-ASCII Content-Length: 2979 ------------------------------------------------------------------- Neural Computing Research Group ------------------------------- Dept of Computer Science and Applied Mathematics Aston University, Birmingham, UK TWO LECTURESHIPS ---------------- * Full details at http://neural-server.aston.ac.uk// * Applications are invited for two Lectureships within the Department of Computer Science and Applied Mathematics. (These posts are roughly comparable to Assistant Professor positions in North America). Candidates are expected to have excellent academic qualifications and a proven record of research. The appointments will be for an initial period of three years, with the possibility of subsequent renewal or transfer to a continuing appointment. Successful candidates will be expected to make a substantial contribution to the research activities of the Group in the area of neural computing, or a related area concerned with advanced information processing. Current research activity focusses on principled approaches to neural computing, and ranges from theoretical foundations to industrial and commercial applications. We would be interested in candidates who can contribute directly to this research programme or who can broaden it into related areas, while maintaining the emphasis on theoretically well-founded research. The successful candidates will also be expected to contribute to the undergraduate and/or postgraduate teaching programmes. Neural Computing Research Group ------------------------------- The Neural Computing Research Group currently comprises the following academic staff Chris Bishop Professor David Lowe Professor David Bounds Professor Richard Rohwer Lecturer Alan Harget Lecturer Ian Nabney Lecturer David Saad Lecturer (arrives 1 August) together with the following Research Fellows Chris Williams Shane Murnion Alan McLachlan Huaihu Zhu a full-time computer support assistant, and eleven postgraduate research students. Conditions of Service --------------------- The appointments will be for an initial period of three years, with the possibility of subsequent renewal or transfer to a continuing appointment. Salaries will be within the lecturer A and B range 14,756 to 25,735, and exceptionally up to 28,756 (UK pounds; these salary scales are currently under review). How to Apply ------------ If you wish to be considered for one of these positions, please send a full CV and publications list, together with the names of 4 referees, to: Professor C M Bishop Neural Computing Research Group Department of Computer Science and Applied Mathematics Aston University Birmingham B4 7ET, U.K. Tel: 021 359 3611 ext. 4270 Fax: 021 333 6215 e-mail: c.m.bishop@aston.ac.uk closing date: 19 May 1995 From tom@csc1.prin.edu Sat Apr 1 13:12:06 1995 Received: from lucy.cs.wisc.edu by sea.cs.wisc.edu; Sat, 1 Apr 95 13:11:55 -0600; AA22616 Received: from TELNET-1.SRV.CS.CMU.EDU by lucy.cs.wisc.edu; Sat, 1 Apr 95 13:08:37 -0600 Received: from TELNET-1.SRV.CS.CMU.EDU by telnet-1.srv.cs.CMU.EDU id aa22572; 31 Mar 95 18:02:43 EST Received: from DST.BOLTZ.CS.CMU.EDU by TELNET-1.SRV.CS.CMU.EDU id aa22570; 31 Mar 95 17:54:01 EST Received: from DST.BOLTZ.CS.CMU.EDU by DST.BOLTZ.CS.CMU.EDU id aa21205; 31 Mar 95 17:53:54 EST Received: from RI.CMU.EDU by B.GP.CS.CMU.EDU id aa10855; 31 Mar 95 17:42:25 EST Received: from spectre.prin.edu by RI.CMU.EDU id aa12149; 31 Mar 95 17:41:24 EST Received: from csc1.prin.edu (csc1.prin.edu [155.106.2.249]) by spectre.prin.edu (8.6.11/8.6.4) with SMTP id QAA22381 for ; Fri, 31 Mar 1995 16:38:13 -0600 Message-Id: <199503312238.QAA22381@spectre.prin.edu> Received: by csc1.prin.edu (1.38.193.4/16.2) id AA00748; Fri, 31 Mar 1995 16:37:50 -0600 Date: Fri, 31 Mar 1995 16:37:50 -0600 From: Tom Fuller Ftp-Host: archive.cis.ohio-state.edu Ftp-File: pub/neuroprose/Thesis/fuller.thesis.ps.Z Apparently-To: Connectionists@cs.cmu.edu The file fuller.thesis.ps.Z is now available for copying from the Neuroprose repository: Supervised Competitive Learning: a technology for pen-based adaptation in real time Thomas H. Fuller, Jr. Computer Science Principia College Elsah, IL 62028 tom@csc1.prin.edu Abstract: The advent of affordable, pen-based computers promises wide application in educational and home settings. In such settings, systems will be regularly employed by a few users (children or students), and occasionally by other users (teachers or parents). The systems must adapt to the writing and gestures of regular users but not lose prior recognition ability. Furthermore, this adaptation must occur in real time not to frustrate or confuse the user, and not to interfere with the task at hand. It must also provide a reliable measure of the likelihood of correct recognition. Supervised Competitive Learning is our technology for the recognition of handwritten symbols. It uses a shifting collection of neural network-based similarity detectors to adapt to the user. We demonstrate that it satisfies the following requirements: 1. Pen-based technology: digitizing display tablet with pen. 2. Low cost: PC-level processor with about 50 MIPS. 3. Wide range of subjects: varying by age, nationality, writing style. 4. Wide range of symbol sets: numerals, alphabetic characters, gestures. 5. Usage: adaptation to regular users; persistent response to occasional users. 6. On-line recognition: both response and adaptation in real time. 7. Self-criticism: reliable measure of likelihood of correct response. 8. Context-free classification: symbol by symbol recognition. SCL successfully recognizes handwritten characters from writers on which it has trained (digits, lowercase, uppercase, and others) at least as well as known current systems (96.5% - 99.2%, depending on character sets). It adapts to its user in real time with a 50 MIPS processor without loss of response to occasional users. Finally, its estimates of its correctness are strongly correlated with actual likelihood of correctness. This is a doctoral dissertation at Washington University in St. Louis. Hardcopies are only available from University Microfilms, Inc. This work was supported by the Kumon Machine Project. ADVISOR: Professor Takayuki Dan Kimura completed December, 1994 Department of Computer Science Washington University Campus Box 1045 One Brookings Drive St. Louis, MO 63130-4899 queries about the work should go to; Thomas H. Fuller, Jr. Computer Science Principia College Elsah, IL 62028 tom@csc1.prin.edu Here's a sample retrieval session: unix> ftp archive.cis.ohio-state.edu Connected to archive.cis.ohio-state.edu. 220 archive FTP server (Version wu-2.4(2) Mon Apr 18 14:41:30 EDT 1994) ready. Name (archive.cis.ohio-state.edu:me): anonymous 331 Guest login ok, send your complete e-mail address as password. Password: me@here.edu 230 Guest login ok, access restrictions apply. Remote system type is UNIX. Using binary mode to transfer files. ftp> cd pub/neuroprose/Thesis 250 CWD command successful. ftp> get fuller.thesis.ps.Z 200 PORT command successful. 150 Opening BINARY mode data connection for fuller.thesis.ps.Z (798510 bytes). 226 Transfer complete. 798510 bytes received in 180 seconds (5 Kbytes/s) ftp> bye 221 Goodbye. unix> uncompress fuller.thesis.ps.Z unix> From rwp@eng.cam.ac.uk Sat Apr 1 13:15:13 1995 Received: from lucy.cs.wisc.edu by sea.cs.wisc.edu; Sat, 1 Apr 95 13:09:55 -0600; AA22607 Received: from TELNET-1.SRV.CS.CMU.EDU by lucy.cs.wisc.edu; Sat, 1 Apr 95 13:06:47 -0600 Received: from TELNET-1.SRV.CS.CMU.EDU by telnet-1.srv.cs.CMU.EDU id ab22415; 31 Mar 95 15:44:53 EST Received: from DST.BOLTZ.CS.CMU.EDU by TELNET-1.SRV.CS.CMU.EDU id ab22410; 31 Mar 95 15:15:06 EST Received: from DST.BOLTZ.CS.CMU.EDU by DST.BOLTZ.CS.CMU.EDU id aa21056; 31 Mar 95 15:14:22 EST Received: from EDRC.CMU.EDU by B.GP.CS.CMU.EDU id aa02595; 31 Mar 95 8:54:33 EST Received: from spanner.eng.cam.ac.uk by EDRC.CMU.EDU id aa08690; 31 Mar 95 8:53:31 EST Received: from dsl.eng.cam.ac.uk (via rwp@dsl.eng.cam.ac.uk [129.169.24.1]) by spanner.eng.cam.ac.uk with ESMTP id OAA27402; Fri, 31 Mar 1995 14:52:12 +0100 Message-Id: <199503311352.12914@dsl.eng.cam.ac.uk> Received: (via rwp@localhost) by dsl.eng.cam.ac.uk id OAA12914; Fri, 31 Mar 1995 14:52:08 +0100 From: Richard Prager Date: Fri, 31 Mar 1995 14:52:07 BST X-Mailer: Mail User's Shell (7.1.1 5/02/90) To: connectionists@cs.cmu.edu, neuron-request@CATTELL.psych.upenn.edu Subject: Cambridge Neural Nets Summer School 1995 +------------------------------------------------------------------+ | FIFTH ANNUAL CAMBRIDGE NEURAL NETWORKS SUMMER SCHOOL | | | | 24-27 July 1995, Emmanuel College, Cambridge, United Kingdom | +------------------------------------------------------------------+ FULLY FUNDED PLACES AVAILABLE FOR EPSRC RESEARCH STUDENTS DISCOUNTED RATES AVAILABLE FOR ACADEMICS SPEAKERS Professor Chris BISHOP - Aston University, Birmingham Dr Herve BOURLARD - Faculte Polytechnique of Mons, Belgium Dr John DAUGMAN - University of Cambridge Professor Geoffrey HINTON - Toronto University Dr Robert JOHNSTON - GEC Hirst Research Centre, Hertfordshire Professor Michael JORDAN - MIT, Boston, Massachusetts Dr Michael LYNCH - Cambridge Neurodynamics Ltd Dr David MACKAY - University of Cambridge Dr Rich SUTTON - GTE Laboratories, Massachusetts Dr Lionel TARASSENKO - University of Oxford PROGRAMME The course will consist of a series of lectures by international experts, interspersed with practical sessions, laboratory tours, poster session, discussion (both formal and informal) and a commercial exhibition. All sessions will be covered by comprehensive course notes and subjects will include: Introduction and overview: Connectionist computing: an introduction and overview Programming a neural network Parallel distributed processing perspective Theory and parallels with conventional algorithms Architectures: Pattern processing and generalisation Bayesian methods and non-linear modelling Reinforcement learning neural networks Multiple expert networks Self organising neural networks Feedback networks for optimization Applications: System identifications Time series predictions Learning forward and inverse dynamical models Control of non-linear dynamical systems using neural networks Artificial and biological vision systems Silicon VLSI neural networks Applications to speech recognition Applications to mobile robotics Financial system modelling Applications in medical diagnostics WHO WILL BENEFIT * Engineers, software specialists and those needing to assess the current potential of neural networks * Technical staff requiring an overview of the subject * Individuals who already have expertise in this area and need to keep abreast of recent developments * Those who have recently entered the field and require a complete perspective of the subject * Researchers and academics working within neural computing areas, as well as all those in industry researching and developing applications Some, although not all, of the lectures will involve graduate level mathematical theory. ACADEMIC DIRECTORS The Summer School will be chaired by members of Cambridge University Engineering Department (CUED) who as members of the Speech, Vision and Robotics Group have current research interests as shown: DR MAHESAN NIRANJAN - speech processing and pattern classification DR RICHARD PRAGER - speech and medical applications DR TONY ROBINSON - recurrent networks and speech processing ACADEMIC SPEAKERS PROFESSOR CHRIS BISHOP, Head of the Neural Computing Research Group at Aston University and Chairman of the Neural Computing Applications Forum, is researching statistical pattern recognition. DR HERVE BOURLARD is with Faculte Polytechnique of Mons, Belgium. He has made many contributions in the area of neural networks and speech recognition. DR JOHN DAUGMAN is Lecturer in Artificial Intelligence in the Computer Laboratory at Cambridge University. His areas of research are computational neuroscience, multi-dimensional signal processing, computer vision, statistical pattern recognition, and biological vision. PROFESSOR GEOFFREY HINTON, Professor of Computer Science and Psychology at the University of Toronto, researches learning, perception and symbol processing in neural networks. He was one of the researchers who introduced the back-propagation algorithm that is now widely used for practical applications. PROFESSOR MICHAEL JORDAN is in the Department of Brain & Cognitive Science at MIT. His contributions to the field include the development of probabilistic methods for learning in modular and hierarchical systems, and the development of methods for applying neural networks to system identification and control problems. DR DAVID MACKAY works on Bayesian methods and non-linear modelling at the Cavendish Laboratory. He obtained his PhD in Computation and Neural Systems at California Institute of Technology. DR LIONEL TARASSENKO is with the Department of Engineering Science at the University of Oxford. His specialisms are robotics and the hardware implementation of neural computing. INDUSTRIAL SPEAKERS DR ROBERT JOHNSTON joined GEC Hirst Research Centre in 1989. His current research is on the applications of non-linear control, with emphasis on fuzzy systems and neural networks. DR MICHAEL LYNCH is Managing Director of Cambridge Neuro-dynamics, a company specialising in the practical application of neural network recognition systems for police, transport and security systems. DR RICH SUTTON is with the Adaptive Systems Department of GTE Laboratories near Boston, Massachusetts. His specialisms are reinforcement learning, planning and animal learning behaviours. LABORATORY TOURS Two afternoon tours will provide the opportunity to observe current research in the field of neural network theory and applications. The Neural Networks Group at CUED have been working in this field since 1984. The Group currently consists of 8 staff and 14 research students. `HANDS-ON' SESSIONS An afternoon practical session will offer the chance to experiment with neural network software and develop an understanding of its strengths and weaknesses. A custom designed environment is available to enable the participants to simulate a variety of problems and explore neural solutions. APPLICATIONS DAY Day 4 will be focused on the applications of neural systems featuring presentations from companies which have exploited connectionist solutions in their businesses. This will give delegates invaluable first hand insight into the technical and practical detail of the transition from research to application. It will present a 'world-class' perspective on the relevance of neural design techniques to commercial and industrial requirements. RESEARCH STUDENTS The Engineering and Physical Sciences Research Council are funding a limited number of places for UK, post-graduate research students. The students benefit from the interaction with a wide range of academics and industrialists and have the opportunity to extend their experience and establish links into industry and other institutions. A poster session will present the current research interests of each student, providing an insight into the work of the connectionist groups in higher education institutions across the UK. VENUE AND ACCOMMODATION The Summer School will be held at Emmanuel College, Cambridge. Emmanuel was founded in 1584 and its attractions include the Wren Chapel and beautiful gardens which delegates may enjoy. The Summer School will take place in the new lecture theatre complex. Emmanuel's city centre location provides easy access to shops and services, is on the main bus route from the Rail Station and is 2 minutes walk from Drummer Street Bus Station, which is served by all national and airport services. Accommodation can be arranged for delegates in single study bedrooms with shared facilities at Emmanuel College for 205 pounds for 4 nights to include bed and breakfast, dinner and a Course Dinner. An additional night's accommodation (bed and breakfast only) on Thursday, 27 July is available at 28 pounds. If you would prefer to make your own arrangements please indicate on the registration form and details of local hotels will be sent to you. SUMMER SCHOOL FEES All fees are payable in advance and include a set of course notes and all day-time refreshments. Days 1-4 (Monday-Thursday) - 775 pounds Academic Discounted Rate: (see qualifying note)* Days 1-4 (Monday-Thursday) - 475 pounds *Academic discounts only available if fees are to be paid by an academic institution. Limited number available - please contact the Course Administrator before applying. REGISTRATIONS For applications for EPSRC fully funded studentships please use the form at the end of this message. Otherwise please contact the Cambridge Programme for Industry: By Email: rjs1008@cus.cam.ac.uk By Post: Registration Administrator, University of Cambridge, Programme for Industry, 1 Trumpington Street, Cambridge, CB2 1QA, UK. By Phone on: +44 (0)1223 302233 By Fax on: +44 (0)1223 301122 All reserved places must be confirmed by returning a registration form to the address shown. Bookings will be confirmed after payment has been received in full. Delegates will only be accepted onto a course if payment has been received in full or an official company order has been received. METHODS OF PAYMENT Payments should be made by: A cheque drawn on a UK bank, VISA or Mastercard/Eurocard, Sterling banker's draft drawn on a UK bank, Crossed international money order, Sterling travellers' cheques Any bank charges arising from international transactions must be met by the delegate. All payments to University of Cambridge must be for the full amount of fees incurred. Personal cheques drawn on banks outside the UK will not be accepted. Please do not send cash. Cheques or orders should be made payable to the 'University of Cambridge-EYA 4814'. CANCELLATIONS Half the registration fee will be returned for bookings cancelled up to one calender month in advance of the course. After this time no fees are returnable. However, substitutions may be made at any time. The Cambridge Programme for Industry reserves the right to pass on any charges levied by a College for cancellation of accommodation and meal bookings. EPSRC FUNDED STUDENTSHIPS A number of Studentships are available for EPSRC-funded, UK registered, post-graduate research students with UK residency. The Studentship covers all course costs and students can attend either Days 1-3 or Days 1-4, with full accommodation packages. Overnight accommodation will not be funded for students resident in or close to Cambridge. Funding is not available for accommodation on Sunday, 23 July, but bed and breakfast in College can be booked at a cost to the student of 28 pounds. Please indicate if you require the extra night's accommodation at the time of application. Wherever possible, students will be funded to their requested level, however it will be necessary to limit the number of students funded for Day 4. Please note that Studentships do not include travel costs. Recipients of 1994 Studentships will not be eligible for 1995 Studentships. HOW TO APPLY FOR AN EPSRC FULLY FUNDED STUDENTSHIP To be considered for a place, please complete the application form below and send a one page summary of current research including how you expect to benefit by attending, a curriculum vitae and a letter of recommendation from your supervisor. The deadline for applications is 19 May 1995. It should be noted that successful applicants will be required to present a poster of their current research or of the research interest of their group. The poster session will be held on Day 1. EPSRC Studentship No. Title (Mr/Miss/Ms) Surname First Name(s) Institution Address Post Code Telephone Fax E-mail: I wish to be considered for an EPSRC Studentship as follows: (please delete those which do not apply) Days 1-3 with 2 nights accommodation package Days 1-3 without accommodation Days 1-4 with 3 nights accommodation package Days 1-4 without accommodation Bed and breakfast only on Sunday, 23 July @ 28 pounds