CNS*1992

The Annual Computational Neuroscience Meeting

July 1992, San Francisco, California

CNS*1992 Abstracts

Author: B.R. Parnas and E.R. Lewis
Department of EECS
University of California
Berkeley, CA 94720

Title: A Neuronal Modeling System for Generating Brainstem Maps of Auditory Space

Abstract: We have developed a simulation system, based on realistic representations of neurons, which exploits massive parallelism to represent the processing performed by the first few levels of the generalized mammalian auditory system. We have modeled the cochlea, auditory nerve (AN), anteroventral cochlear nucleus (AVCN) and medial superior olivary nucleus (MSO). These structures produce a variety of maps of auditory space which are used by further processing centers to extract the cues used in auditory perception. The centers we have modeled thus far play a role in binaural signal processing.
Author: K. Pawelzik, H.-U. Bauer, T. Geisel
Institut fur Theoretische Physik and SFB Nichtlineare Dynamik
Universitat Frankfurt, Robert-Mayer-Str. 8-10
W-6000 Frankfurt/Main 11, Fed. Rep. of Germany

Title: Switching Between Predictable and Unpredictable States in Data from Cat Visual Cortex

Abstract: Visual inspection of data from cat visual cortex reveals, that the recently discovered 40-Hz-oscillations do not consist of a permanent oscillatory signal during the whole period of stimulation, but instead display interchanging stochastic and regular phases, both of varying duration times. In order to base this observation on firm statistical grounds, we assume that the signal should be predictable during regular epochs, unpredictable during the irregular epochs. Employing the method of time-resolved mutual information, first we calculate the time-resolved predictability within the LFP-signal from a single electrode. Periods of high temporal coherence can clearly be discrminated from periods of stochastic behaviour. The analysis of the spatial coherence with the same method yields, that synchronous activity between two electrodes is largely restricted to episodes of high temporal coherence.
Author: Pellionisz, A.J. and Bloedel J.R. NASA
Ames Research Center VRF 242-3 CA
and Barrow Neurological Institute,
350 West Thomas Rd, Phoenix, AZ 85013

Title: Geometry Intrinsic to Population Responses of Cerebellar Purkinje Cells as Revealed by Neurocomputer Analysis of Multi-Unit Recordings from Biological Neural Nets

Abstract: A most important basic research goal of neurocomputing is to identify, from biological neural nets, the mathematics intrinsic to neural net function. From up to ten cerebellar Purkinje cells of locomotory cats, multi-electrode recordings were obtained and the geometry of the functional manifold was calculated. The covariant metric tensor was calculated by cross-correlation analysis and the contravariant metric tensor was obtained by its Moore-Penrose inverse. For adequate analysis of massively parallel data, a transputer-based neurocomputer was built. Results demonstrate the non-Euclidean neural geometry and its modifiability with learning. Identification of geometry permits decoding, of external invariants (distances) from neural firings.
Author: Eric O. Postma, H. Jaap van den Herik and Patrick T. W. Hudson
Department of Computer Science
Faculty of General Sciences (FdAW)
University of Limburg
PO Box 616, 6200 MD Maastricht
The Netherlands

Title: The Gating Lattice: A Neural Substrate for Dynamic Gating

Abstract: The Gating Lattice is proposed as a basic neural building block for dynamic gating of patterns between cortical areas. Dynamic gating involves the selection of contiguous patterns from an input stage, which are then routed via one or more intermediate stages towards an output stage. A Gating Lattice consists of many distributed processors whose local interactions ensure appropriate selection, and whose global behavior can be described in terms of statistical mechanics. Combined with a modification of the shifter-circuit architecture (Anderson and Van Essen, 1987) it represents a model for attentional selection in the occipitotemporal pathway.
Author: Brian Ringer and Josef Skrzypek
Machine Perception Laboratory
Computer Science Department
University of California, Los Angeles
Los Angeles, CA 90024

Title: A Neural Model of Illusory Contour Perception

Abstract: Illusory contours, which are perceived in absence of luminance contrast data, play an important role in visual processing. However, it is not clear what kind and how much information is necessary for the perception of this phenomenon. Most of the available explanations assume that contrast information on both sides of a gap is sufflcient to complete an illusory contour. Using a computer simulation we evaluated hypothesized explanation by making explicit physiological and psychophysical constraints. Our results indicate that a solution to illusory contour perception may lie with a recurrent network which integrates information about illusory surfaces (brightness and binocularity) with available contour information (luminance contrast).
Author: Thomas B. Schillen and Peter Konig
Max-Planck-Institut fur Hirnforschung
Deutschordenstrasse 46, 6000 Frankfurt 71, FRG

Title: Temporal Structure can solve the Binding Problem for Multiple Feature Domains

Abstract: We investigate a solution to the binding problem in visual processing using temporal structure in a neuronal network. First, we demonstrate binding and segregation of assemblies by synchronizing and desynchronizing connections for a single feature representation. Then, we extend this model to multiple feature domains where each member of an distributed assembly is specific for its single particular feature expression only. This avoids the combinatorial explosion of multispecific cardinal cells. Besides binding by synchronization of temporal structure of distributed neuronal responses, our simulations demonstrate also for these examples that synchronization has to be complemented by a means for stimulus-dependent desynchronization.
Author: By J.S. Schwaber, J.F.R. Paton, R.F. Rogers and E.B. Graves
Neural Computation Group,
DuPont Experimental Station E352, Dupont Company,
Wilmington, DE 19880-0352, U.S.A.
and Institute for Neuroscience, University of Pennsylvania,
Philadelphia, PA 19104

Title: Modeling neuronal dynamics predicts responses in the rat baroreflex

Abstract: Computer models were constructed for neurons recorded in vitro in the cardiovascular NTS of the rat. These neurons showed time- and voltage-dependent spike discharge patterns and subthreshold voltage trajectories when they were intracellularly injected with constant current pulses. Models of membrane ion channels were constructed and used to produce and match this dynamic behav- ior. Network models composed of these neurons and reflecting baroreceptor reflex organization in the NTS were used to explore systems level computation within the reflex. Highly non-linear and unpredicted activity was observed in simulations. However, some of these simulation results were taken as hyptheses and have subsequently been observed experimentally anatomically and in vivo.
Author: Hagai Agmon-Snir and Idan Segev,
Dept. Neurobiol. Inst. Life Sciences,
Hebrew Univ. Jerusalem, 91904, Israel

Title: Dendritic delay

Abstract: Neurons behave like delay lines. A significant delay is contributed by the cable properties of the dendrites as synaptic potentials propagate from their sites of origin towards the spike initiation zone. We introduce a novel analytical approach for calculating signal delay in arbitrary passive structures. Dendritic delay is defined as the difference between the center of gravity (the centroid) of the transient current input and the center of gravity of the resultant voltage transient, measured at any point in the dendritic tree. Several general theorems on the effect of various biophysical and geometrical parameters on the delay are proven. Specific examples for the delay in different geometries, ranging from an isopotential soma to complicated trees, are given. We discuss the implications of the results for the computations that may be performed by nerve cells.
Author: Adam Prugel-Bennett and Jonathan L. Shapiro
Department of Computer Science,
Oxford Road,
Manchester

Title: Unsupervised Hebbian Learning and the Shape of the Neuron Activation Function

Abstract: A model describing unsupervised Hebbian learning in a single neuron is presented. The dynamical equations are solved directly in a few simple cases, methods of statistical mechanics are used to compute the probability distribution for the stationary solutions for random patterns, and simulations are presented. The processing function which the neuron learns is strongly dependent on the shape of the activation function. When the activation function is linear, the neuron learns to compute a statistical property of the collection of input patterns. When it is sufficiently nonlinear, the neuron forms a grandmother cell representation of one of the patterns.
Author: Josef Skrzypek and George Wu
Machine Perception Laboratory
Computer Science Department
University of California, Los Angeles
Los Angeles, CA 90024

Title: Invariant Contrast Adaptation in the Outer Plexiform Layer of the Primate Retina

Abstract: Recent anatomical and physiological results suggest that luminance contrast adaptation in the primate retina is localized to synaptic interactions of the outer plexiform layer(OPL). We present a quantitative study of luminance contrast adaptation using a computer model of the primate OPL derived from known anatomy and physiology. Results demonstrate that simple network interactions can implement invariant contrast adaptation while maintaining high luminance contrast sensitivity over 7 log units of retinal illuminance. Furthermore, simulation results predict behavior consistent with known psychophysics of primate foveal photopic vision. As far as we know this is the first quantitative demonstration of compatibility of primate OPL anatomy, physiology and psychophysics.
Author: Diana Smetters
Department of Brain and Cognitive Science
43 Carleton St., E25-618
MIT
Cambridge, MA 02139

Title: A Model of a Thalamic Relay Cell Incorporating Voltage-Clamp Data on Ionic Conductances

Abstract: We have constructed a model of a thalamic relay cell based on voltage clamp data obtained in thalamic slices and isolated relay cells. This model replicates most, but not all, of the behavior found in current-clamp recordings from these cells. Analysis of the model provides insight into the role of the various voltage-dependent conductances in the response of these cells to current injection and synaptic input. Additionally, the model provides insight into the insufficiency of traditional methods of measuring passive parameters, and may provide a tool for better understanding of voltage-clamp results.
Author: Jarnes A. Smith
Center For Biological Information Processing,
E25-201 MIT Cambridge, MA 02139,
and Norberto M. Grzywacz
Smith-Kettlewell Institute,
2232 Webster St. San Francisco, CA 94115.

Title: A local model for transparent motions based on spatiotemporal filters

Abstract: Psychophysical studies suggest that humans perceive moving sine plaids as fused when the underlying component gratings are of sufficiently different speed, orientation, and contrast (Adelson & Movshon 1982; Welch, 1990). Other experiments indicate that for short presentation times, the velocities of fused type II sine plaids are not always consistent with the intersection of constraints. We propose a gradient inhibition model that explains these phenomena using only local intensity information. Computational results show a contrast dependent transition for splitting/fusion of sine plaid images. This model suggests that transparency information can be locally computed and subsequently fed into global vision modules.
Author: William R. Softky and Christof Koch
216-76 California Institute of Technology
Pasadena, CA 91125

Title: Cortical Cells Should Fire Regularly, But Do Not

Abstract: The standard theory of a leaky integrator with stochastic spike input predicts that cortical cells should fire regularly. We tested interspike-interval histograms from awake, behaving macaque visual cortex (Vl and MT), and found high levels of variability (CV > 0.5) characteristic of a nearly random (Poisson) process. A simple integrate-and-fire model, using accepted biophysical parameters, fails by more than a factor of ten to account for the high CV. We also simulated a biophysically detailed compartment-model of an anatomically reconstructed and physiologically characterized layer V pyramidal cell; again, at high firing rate, CV values are low, in disagreement with the data. Only a few situations could account for this discrepency: very large EPSP's, a very short membrane time constant (Tm < 0.3 msec), or highly correlated (e.g. synchronized) input. Our analysis suggests that cortical computation may occur at a time-scale much faster than previously realized.
Author: Michael D. Speight and James Bower
Division of Biology 216-76
California Institute of Technology
Pasadena, CA91125

Title: Using Parallel Supercomputers for Computational Neurobiology

Abstract: We present our experiences in providing the GENESIS neural simulation system on two different parallel supercomputers one of which is currently the world's fastest computer (Intel Touchstone Delta). Two distinct methods of utilising the power of these novel architecture machines are presented. One method (Task Farming) scales by order 'N' with the number of nodes used, and is particularly useful for examining large parameter spaces. The other method (Distributed Modelling with Internode Communication) allows construction of neural models of a size and computational complexity not hitherto able to be achieved using conventional supercomputing platforms.
Author: M.A. Srinivasan
Research Lab of Electronics, MIT
Cambridge, MA 02139

Title: Computations in tactile sensing

Abstract: Our tactile sensation is the culmination of a series of events : Physical contact with an object causes mechanical loading on tlle skin surface and results in distortions of mechanoreceptors;; the receptors, in turn, respond with electrical impulse trains hlat are subsequently processed by the nervous system. In this paper, we present detection of slip, microtexture, shape, alld compliance as examples of computations in tactile sensillg. We draw upon results from our experiments on the biomechanics, neurophysiology, and psychophysics of tactile sense, as well as theoretical analyses employing tlle mechanics of deformable media.
Author: Adam F. Strassber and Louis J. DeFelice
Computation and Neural Systems Program, California Institute of
Technology, Pasadena, CA
Caltech 216-76, Pasadena, CA 91125

Title: Limitations of the Hodgkin-Huxley Formalism: Effects of Single Channel Kinetics upon Transmembrane Voltage Dynamics

Abstract: The effects of random fluctuations of single ion channels upon transmembrane voltage dynamics are investigated with the example of the giant axon of the squid Loligo. A simulation using the continuous, deterministic Hodgkin-Huxley equations is compared to a simulation using discrete, stochastic ion channel populations modeled as Markov processes. The relative convergence and divergence of the membrane behavior predicted by the two alternative simulations is compared. It is demonstrated that random microscopic behavior, such as single channel fluctuations,can producerandom macrosopic behavior, such as action potentials, under many common biophysical conditions. The various parameters contributing to the amplification of this channel noise are analyzed and estimates for the noise sensitivity of a given membrane are derived. Many interesting biological adaptations have evolved to either exploit or suppress thischannel noise. These biological regimes are surveyed and their computational significance is considered.
Author: Humbert Suarez, Christof Koch, and Rodney Douglas
Computation and Neural Systems Program,
California Institute of Technology,
Pasadena CA 91125, USA.
Anatomical Neuropharmacology Unit, University of Oxford, Oxford,
United Kingdom.

Title: A Model of Direction Selectivity in Visual Cortex Using Massive Intracortical Connections

Abstract: Almost all models of orientation and direction-selectivity in visual cortex are based on feed-forward connection schemes, where geniculate input provides all excitation to both pyramidal and inhibitory neurons. The latter neurons then veto the response of the former for non-optimal stimuli. However, the majority of the synaptic input on any cortical cell is provided by other cortical cells. Based on the canonical microcircuit of Douglas \& Martin (1991), we simulate here this massive excitatory and inhibitory synaptic convergence in order to explain a number of puzzling features about direction selective simple cells. In this model, weak geniculate input is dramatically amplified by the action of intracortical excitation, while inhibition has a dual role: (i) to prevent the early geniculate-induced excitation in the null direction and (ii) to serve as gain-control element.
Author: W. Edward Sullivan
Department of Ecology and Evolutionary Biology and Program in Neuroscience
Princeton University
Princeton, NJ 08544

Title: The highs and lows of horizontal sound localization: The biophysics of Structure-Function correlations in nucleus laminaris.

Abstract: Computer simulations designed to study horizontal sound localization suggest that morphological differences between homologous neurons in barn owls and other birds are related to the range of frequencies employed and to the biophysical algorithm used for "coincidence detection". Barn owls measure interaural phase differences from 3 - 9 kHz. The cells performing this function are a-dendritic and receive many synapses from each side. In simulation, coherent phase-locked inputs produce a synaptic modulation at the stimulus frequency that is proportional to interaural phase difference, being absent for out-of-phase binaural inputs. Spike generators selective for rapid voltage modulations can thus be used for coincidence detection. Modifications of the Hodgkin-Huxley model were found that enable such selectivity. In contrast, chickens use lower frequencies and have coincidence detectors with long bipolar dendrites and long, thin axon initial segments. These can provide greater temporal discrimination at the expense of high frequency responsiveness. High and low frequency coincidence detection may therefore be achieved by different anatomical and biophysical means.
Author: David C. Tam
Division of Neuroscience
Baylor College of Medicine
Houston, TX 77030

Title: A multi-Neuronal Vectorial Phase-space analysis for detecting Dynamical interactions in firing patterns of Biological Neural

Abstract: A novel vectorial statistic is introduced to detect temporally correlated firing patterns in a network of n neurons. A "cross-interval" vector is used to establish the temporal relationship of cross-intervals between firings in neurons. The resultant vectorial sum of these vectors provides a statistical measure of an n-tuple correlation among all n neurons in the network in contrast to the pair-wise correlation limitation in cross-correlation analysis. The normalized resultant vectors not only capture and reduce an O(n3) combinatorial correlation to an O(n) vectorial statistic but also provide quantitative descriptions of dynamical interactions from the trajectories and clusters of these vectors in the phase-plane. It provides another unique technique to analyze the interactions among multiple processes in a nonlinear dynamical system.
Author: F.E. Theunissen, S. N. Gozani and J.P. Miller.
Dept. of Mol. and Cell
Biology. University of California, Berkeley. CA 94720.

Title: Representation of sensory information in the cricket cercal sensory system: an information theoretic analysis.

Abstract: The activity patterns of primary sensory interneurons in the cricket cercal sensory system encode information about the direction and velocity of air current stimuli in the animal's immediate environment. The statistical principles of information theory were used to calculate the maximum directional accuracy attainable from the response ensemble of 8 interneurons in this system. These calculations of encoding accuracy took into account the temporal patterns of the spike train patterns, and a canonical discriminant analysis was used to reduce the number of dimensions of the response to a few significant ones. The response of the system in the canonical coordinates could be fitted by sine functions of 360 and 180 degree periods.
Author: John H. Thompson and James M. Bower

Title: Electrophysiological dissection of the three excitatory inputs to cerebellar Purkinje cells

Abstract: Experiments in our and other laboratories have described the spatial organization of tactile mossy fiber projections to the granule cell layer of crus IIa of rat cerebellar cortex. Using high density micro-mapping techniques, it has been shown that the perioral regions are represented by a fractured somatotopic map. That is, there are patches within whose boundaries a region of the rat's face is somatotopically mapped. At patch boundaries, however, there are discontinuities after which a new, not necessarily peripherally adjoining region is mapped. The granule cells which receive these inputs project to the molecular layer which contains the dendrites of the Purkinje cells. In this layer the granule cell axons bifurcate forming the parallel fibers which run parallel to each other for several millimeters across patch boundaries contacting numerous Purkinje cells along their length. Bower and Woolston (1983) found that the shortest latency tactile projections to the Purkinje cells in this region were spatially congruent to those in the granule cell layer. That is, activated Purkinje cells were directly over activated granule cells. This suggested that the ascending portion of the granule cell axon (which synapses on Purkinje cells on the way up) was responsible for this short latency response.

In these experiments we record from a Purkinje cell while stimulating perioral regions projecting to the underlying granule cells, constructing a peri-stimulus time histogram (PST) from the spikes recorded from 300 trials. Then, the stimulus is moved to other regions which project to patches in the same folium but removed from the Purkinje cell. In separate trials, both regions which send parallel fibers to the Purkinje cell and some which don't are stimulated. Comparing the PSTs from the various trials, we conclude that the short latency excitatory response is mediated by the input from the ascending portion of the granule cell axon only. There is a prolonged, longer latency response starting approximately 40 ms after the stimulus and lasting for 100 to 400 ms which is mediated by the parallel fibers. There is little or no response to stimuli in regions not connected by the parallel fibers. Lastly, we are able to discriminate between the simple spikes in Purkinje cells generated by the two granule cell inputs and the complex spikes generated by the other input pathway, the climbing fibers from the inferior olive. We have not been able to statistically link the climbing fiber pathway to the generation of either the short or long latency responses.

The results of these experiments are studied in the context of a new interpretation of the functional significance of cerebellar cortical architecture. Purkinje cells respond to specific information relayed via the ascending branch in the context of more subtle modulation provided by the parallel fibers. This allows a new understanding of the fractured somatotopic organization of mossy fiber projections.
Author: Bryan J. Travis
MS-F66S
Los Alamos National LaboratoIy
Los Alamos, NM 87545

Title: A Computational Model of the Cat Subcortical Auditory System

Abstract: A literal computational model of the cat subcortical auditory system, from the cochlea to the medial geniculate body of the thalamus is described. The model includes subunits representing the basilar membrane/hair cell system of the cochlea, parts of the cochlear nucleus, the central nucleus of the inferior colliculus and the ventral division of the medial geniculate body in the thalamus. Each subunit consists of model neurons whose structure, dynamics and connectivity are based as much as possible on what is known of the cat auditory system. The goal of this modeling effort is to shed light on the role of (1) the convergent/divergent projections seen between each stage of the auditory system and (2) the efferent projections sent from higher stages to lower.
Author: Alessandro Treves and L F Abbott
Department of Experimental Psychology,
University of Oxford
South Parks Road, Oxford OX1 3UD, U K

Title: Lateral Processing: the Time it Takes

Abstract: A novel analytical approach is presented, which allows studying the dynamics of reciprocally connected networks of integrate-and-fire model neurons. Synaptic interactions are modelled at the level of conductance changes, and firing rate adaptation is induced by a spike-activated intrinsic potassium conductance. One obtains both the attractor states of the network and the full spectrum of time constants of the transients leading to those attractor states. Using parameters appropriate to neocortical pyramidal cells, one can estimate, from the slowest transients, the time required for lateral processing, and compare with new experimental data on the time course of single-cell responses in behaving primates.
Author: P. S. Ulinski and L. J. Larson-Prior,
Department of Organismal Biology and Anatomy,
University of Chicago,
Chicago, IL 60637.

Title: Analysis of the distribution of geniculate afferents upon pyramidal neurons in turtle visual cortex

Abstract: Pyramidal neurons in the lateral visual cortex receive geniculate afferents upon their proximal dendrites, while those in the medial part receive geniculate afferents upon their distal dendrites. Geniculate EPSPs recorded in vitro from the somata of lateral pyramidal neurons consist of 2-3 unitary EPSPs distributed over a time window of 25 msec while medially located pyramidal neurons exhibit 5 or more unitary EPSPs over 100 msec The very different waveforms generated in lateral versus medial pyramidal neurons can be understood by the differences in the distribution of geniculate afferents upon their dendritic arbors. Supported by PHS Grant EY 68352
Author: Michael C. Vanier and James M. Bower
Division of Biology 216-76
California Institute of Technology
Pasadena, California 91125

Title: Noradrenergic modulation of synaptic transmission in piriform cortex

Abstract: We have demonstrated that bath-applied norepinephrine in doses ranging from 10 - 50 uM can cause a long-lasting increase in the strength of intrinsic excitatory synaptic evoked potentials, which may imply that it is serving as a "learning switch" to increase plasticity during behaviorally relevant situations. Norepinephrine treatment initially causes a rapid and dose-dependent suppression of synaptic transmission at excitatory intrinsic fiber synapses in rat piriform cortex, similar to but weaker than the response to the acetylcholine agonist carbachol reported in an earlier study (Hasselmo and Bower 1991). Unlike carbachol, in roughly half the cases investigated large increases in excitatory evoked potentials occurred upon withdrawal of norepinephrine from the medium, without using tetanizing stimuli. We are incorporating these findings into our realistic models of associative memory function in piriform cortex.
Author: F. Bini Verona M. Mastroianni S. Russo
Dipartimento di Informatica e Sistemistica
Universita di Napoli
Via Claudio 21, Napoli, I-80125 Italy
and
G. Ventre
International Computer Science Institute
1947 Center Street, Suite 600
Berkeley, California, 94704-1105

Title: Neural Network Simulation in a CSP Language for Multicomputers

Abstract: In this paper we investigate the simulation of neural networks on distributed-memory, message-passing multiprocessor systems, by means of a CSP-like language and its own programming environment designed and implemented at the Universita di Napoli. Many interesting features make this language suitable for simulation and fast prototyping of neural network models. Among them there are generality, modularity and ease of use. In addition, its implementation on distributed systems allows the development of scalable and portable applications and assures a very promising level of efficiency.
Author: T. Wadden, S. Grillnert, A. Lansner & T. Matsushimat
SANSÑStudies of Artificial Neural Systems
Department of Numerical Analysis and Computing Science
Royal Institute of Technology, S-100 44 Stockholm, Sweden

Title: Realistic Simulation of Undulatory Locomotion A Trailing Oscillator Hypothesis

Abstract: In vertebrates like eel and lamprey a laterally directed wave is propagated down the body, pushing the animal forwards through the water. Whereas the function of the segmental neuronal network underlying locomotion (lamprey) is understood, the intersegmental coordination has remained enigmatic. Each spinal segment is connected by excitatory interneurons from neighbouring segments. Our model (7 to 20 segments) of this experimentally established network generates intersegmental coordination. Furthermore the phase lag can be graded and made to change sign, (enabling forward and backward swimming), by adding extra excitation only to one segment in the rostral or caudal part of the oscillator chain.
Author: DeLiang Wang
Department of Computer and Information Science
and Center for Cognitive Science
Ohio State University
2036 Neil Avenue
Columbus, OH 43210-1277, USA

Title: Modeling Stimulus Specific Habituation: The Role of the Primordial Hippocampus

Abstract: We present a neural model for the organization and neural dynamics of the medial pallium, the toad's homolog of mammalian hippocampus. A neural mechanism, called cumulative shrinking, is proposed for mapping temporal responses into a form of population coding referenced by spatial positions. Synaptic plasticity is modeled as an interaction of two dynamic processes which simulates both short-term and long-term memory. Successful modeling allows us to provide an account of the neural mechanisms of stimulus specific habituation. A set of model predictions is presented, conceming mechanisms of habituation and cellular organization of the medial pallium.
Author: X.-J. Wang and J. Rinzel
Mathematical Research Branch, NIDDK,
National Institutes of Health
Bldg. 31, Rm. 4B-54, Bethesda, MD 20892
Department of Mathematics and the James Franck Institute
University of Chicago, 5734 S. University Av.,
Chicago, IL 60637

Title: Synchronization Among Inhibitory Model Neurons: Interplay Between Rebound Excitation and Synaptic Kinetics

Abstract: It is well known that mutual inhibition in a neuronal network can lead to alternating rhythmic patterns of electrical activity. Here, in contrast, we show with a theoretical model that mutually inhibitory cells, capable of post-inhibitory rebound, can be synchronized to zero phase difference, provided that synaptic activation decays slowly enough to outlast the presynaptic excitation. This study is motivated by consideration of a hypothesis on synchronizing mechanisms for the thalamocortical 10 Hz spindle rhythms, that the reticular thalamic nucleus (RTN) plays a role of pacemaker or synchronizer of spindles in the thalamus. Our results suggest that synchronization among GABAergic RTN cells can be mediated by GABA-B postsynaptic receptors which possess the required kinetic properties.
Author: Xin Wang Edward K. Blum
Department of Mathematics, University of Southern California
Los Angeles, California 90089-1113

Title: Modeling Synchronization of Oscillations in Visual and Olfactory Cortex by Coupled Discrete-Time Limit-Cycle and Chaotic Neural Network Oscillators

Abstract: The recent neurophysiological experiments revealed that the spatial synchronization of neural activity across a population of oscillatory neurons exist in the sensory systems. A discrete-time model of coupled limit-cycle and chaotic neural network oscillators is constructed to investigate the coupling structures of the oscillators that are capable of generating the synchronization in neural activity similar to that observed in the experiments.Simulations on the responses of one- and two-dimensional coupled limit-cycle oscillators to two short bar stimuli of varying gap distance and on the phase transitions of coexisting chaotic attractors of coupled chaotic oscillators to different coded inputs are performed.
Author: Lloyd Watts
Mail Stop 116-81
Physics of Computation Laboratory
California Institute of Technology
Pasadena, CA 91125

Title: Designing Networks of Spiking Silicon Neurons and Synapses

Abstract: Carver Mead has pioneered the use of analog VLSI for building silicon systems based on neurobiological principles. As the computational and communication benefits of spiking neurons become more appreciated, there is a growing interest in increasingly realistic "silicon neuron" circuits. Many groups are developing silicon synapses, which, combined with the silicon neurons, will allow the creation of large biologically realistic networks of spiking neurons on a single piece of silicon. As these networks become more and more sophisticated, there will be a need for simulation tools that are appropriate for the circuit primitives. This paper describes simple but realistic silicon neurons, synapses, and a fast event-driven simulator optimized for networks of these analog silicon neurons and synapses. The silicon neuron and synapse circuits have been fabricated and successfully tested, both individually and in small networks. Examples are presented from working chips and compared to results from the simulator.
Author: Thelma L. Williams
Physiology Department,
St.George's Hospital Medical
School, London SW17 ORE, U.K.

Title: Entrainment of the Lampraey locomotor CPG: analysis and simulation

Abstract: Entrainment of a chain of neural oscillators by an external oscillator coupled at one end has been simulated, using connectionist networks. The unit oscillator of the chain consists of synaptically coupled cells postulated to form the basis of the central pattern generator for locomotion in the lamprey. The frequency range over which one-to-one entrainment occurs has been determined for different forms of intersegmental coupling. The dependence of this entrainment range on the characteristics of coupling, and the effect of entrainment on intersegmental phase lags have been investigated and compared with the predictions of mathematical analysis of oscillator chains.
Author: F. Wolf, H.-U. Bauer, T. Geisel
Inst. fur Theor. Physik and SFB "Nichtlineare Dynamik",
Univ. Frankfurt
Postfach 11 19 32, D-6000 Frankfurt/Main 11,
Fed. Rep. of Germany

Title: Field Discontinuities and Islands in Visual Cortical Maps as a Self-Organized Phenomenon

Abstract: We present an explanation for the formation of field discontinuities (FDs) in the visual cortical maps of various mammals. Simulating retinocortical map formation with selforganizing feature maps (SFMs), we find a critical influence of the geometry of target area boundaries on the global structure of the map. One or more FDs are formed as the elongation of the cortical area increases. This is in good agreement with the general observation that second order transformations of the visual field are found in elongated areas only. The observation of islands in the visual field representation of cat A18/19 is explained by the occurrence of more than one FD. Generally the maps show a local increase of the areal magnification factor near the origin of the FD. This prediction might be accessible to future experiments.
Author: Haiyun Zhang and Gilles Laurent,
California Institute of Technology,
Biology Division, CNS Program,
139-74, Pasadena,CA 91125.

Title: Spatial and temporal integration by locust nonspiking interneurons: Voltage-dependent electrical geometry.

Abstract: We constructed a simplified computer model of a locust nonspiking interneuron, using morphological features from intracellular cobalt and HRP fills, and biophysical parameters determined from voltage-clamp studies. The model has 56 compartments, a leak and 2 voltage-activated K conductances with different kinetics of activation and inactivation. These active conductances are known from current-clamp experiments to lead to a ten-fold change in membrane time constant over the normal operating range of the interneurons. We ran numerical simulation experiments to study the changes in apparent space constant with membrane voltage. We found that the electrical geometry of a nonspiking interneuron can be dramatically altered by small changes in membrane voltage, leading to compartmentalization at the more positive potentials.