CNS*1992

The Annual Computational Neuroscience Meeting

July 1992, San Francisco, California

CNS*1992 Abstracts

Author: Subutai Ahmad
International Computer Science Institute
and
Siemens Corporate Research
Siemens AG, ZFE ST SN6
Otto-Hahn Ring 6,
8000 Munich 83, Germany

Title: A Model of Human Visual Attention

Abstract: Visual attention is the ability to dynamically restrict processing to a subset of the visual field. Such a mechanism is necessary to efficiently perform many intermediate level visual tasks. This presentation describes VISIT a novel neural network model of covert visual attention. The current system models the search for target objects in scenes containing multiple distractors. This is a natural task for people, it is studied extensively by psychologists, and it requires attention. The network's behavior closely matches the known psychophysical data on visual search and visual attention. VISIT also matches much of the physiological data on attention and provides a novel view of the functionality of a number of visual areas.
Author: Ashley M. Aitken
Artificial Intelligence Laboratory, Schools of Electrical Engineering and Computer Science & Engineering,
The University of New South Wales,
Box 1, PO Kensington, N.S.W. 2033, AUSTRALIA.

Title: Overview of the Valentino Computational Neuroscience Workbench for Simulation of Neural Systems

Abstract: We present an innovative simulator system designed to facilitate research in computational neuroscience. The Valentino Computational Neuroscience Workbench (CNW) incorporates an object-oriented design and a neuroscience workbench metaphor wherein a number of independent tools manipulate a shared persistent simulation structure. Central to the system is a novel tool for configuring structured neural networks using a graphical user-interface and powerful interconnection selection criteria and schema. Together these make the Valentino CNW powerful, flexible, extendible and, yet, still easy-to-use - unlike some other presently available simulators.
Author: Thomas J. Anastasio
University of Illinois
Beckman Institute

Title: Recurrent backpropagation models of the Vestibulo-Ocular Reflex provide experimentally testable predictions

Abstract: Previous, static backpropagation models of the vestibulo-
oculomotor system were able to capture the distributed aspects of
eye-movement command representation by brainstem neurons.
However, these models do not readily offer testable predictions.
More recently, recurrent backpropagation models have been used to
study the dynamic and nonlinear features of the vestibulo-ocular
reflex (VOR). The dynamic models make clear predictions concern-
ing the behavior of VOR neurons following lesions. Some of the
predictions from the recurrent backpropagation models differ in
critical ways from those derived from analytical models of VOR.
The testability of the recurrent models encourages a continued
dialog between theory and experiment.
Author: Charles H. Anderson~. Bruno Olshausen2, David Van Essen3
Jet Propulsion Laboratory1
Computation and Neural Systems Program2
Division of Biology3
California Institute of Technology
Pasadena, CA 91125

Title: Neural Routing Circuits in Primate Visual Cortex

Abstract: Directed visual attention can be modeled by neural routing circuits that provide high-level areas of the visual cortex with dynamic access to retinal sensory inputs. This approach has many computational advantages over the more static, feedforward models of vision which construct increasingly more complex receptive fields as one moves upward through the hierarchy of cortical areas. Here, we provide details on the structure of neural routing circuits and their control as illustrated by interactive computer programs. Neural substrates for routing and control are proposed and physiological support for the model is discussed.
Author: N. Arad & E. Ruppin & Y. Yeshurun
School of Mathematical Sciences
Sackler Faculty of Exact Sciences
Tel Aviv University
69978, Tel Aviv, Israel

Title: Dopaminergic Induced Multi-stable States in Neural Networks

Abstract: The dynamical behaviour of neural networks is typically considered to be controlled by the synaptic weights and by the neurons' thresholds. Recently, it was suggested that the network's activity can be modified by solely modulating the neuronal activation function. We examine the effect of modulatory neurotransmitter on the behavior of a simple excitatory neural network model by varying the neuronal gain. The network passes through several qualitatively distinct regions of activity, traversing from a uni-steady state to a multi-steady state region, and back to a region of a small number of steady states. Our results may shed light on the possible regulatory role of dopamine in the cortico-striato-pallido-thalamic loop, and the effects of its malfunction in the pathogenesis of movement related disorders.
Author: Christopher Assad, Brian Rasnow, James M. Bower

Title: Numerical simulations of the electric organ discharge of weakly electric fish.

Abstract: A model of a weakly electric fish was constructed with data taken from Apteronotus leptorhynchus, and the electric organ discharge was simulated using boundary element and finite element methods. Maps of the electric potential measured around a live fish were used to calibrate the model parameters and test the results. Initial results suggest the fish's body position has a significant effect on the electrical images detected by the fish. Our goal is to quantitatively simulate various sequences of electrosensory input being sent to the central nervous system as the result of the fish's exploratory behavior.
Author: H. Axelrad
Lab. of Neurophysiology- Fac. Medecine Pitie- Paris13 - France
Dept. of Theoretical Physics - CEA - Saclay - France

Title: A study of computation in a simple Neuronal Network: experimental recording and theoretical simulation of the immature rat cerebellar cortex

Abstract: Experimental and theoretical analysis of the 5-7 days old rat cerebellar cortex (Cbcx) was undertaken to get some insight on the role of inhibition in modulating the ongoing activity and the spatial distribution of functionally correlated neurons. Pairs of Purkinje cells (PC) were recorded in vivo before and after uncoupling the inhibitory links (mediated by their recurrent collaterals) via superfusion of biccuculin. Statistical techniques were used to assess the characteristics and correlation of the spike trains. A new tool is presented which computes an index of informational entropy. Experimental results are compared with simulation results from a biologically accurate model of this structure. It appears that inhibition: 1- significantly changes the spike trains data; 2- it restricts the space of states in which the network can evolve and 3-it leads to a spatio-functional segregation of correlated PCs. The limitations of CNS models will be discussed.
Author: Wyeth Bair
Computation and Neural Systems Program,
California Institute of Technology, Pasadena,CA.

Title: Power Spectrum Analysis of MT Neurons from the Awake Monkey

Abstract: We analyzed extracellular spike trains recorded from single cells in area MT in rhesus monkeys which performed a demanding motion discrimination task (Newsome, Britten & Movshon, 1989). In particular, we computed the power spectra of spike trains from 2 sec long trials of visual stimulation during which the monkey held fixation. Most cells have a relatively flat spectrum with a dip at low frequencies, indicative of a Poisson process with a refractory period. 10% of cells have a substantial peak in the spectrum between 25-50~z, but the presence and strength of this peak does not correlate in any way with the know behavior of the monkey (e.g. correct or incorrect decision). However, there exists a very strong correlation between cells that discharge 2-4 spikes within 37msec (burst cells) and cells that have a peak in the power spectrum. The simplest interpretation of this data is that a small fraction of cells randomly and independently of the stimulus fire in bursts followed by a refractory period, giving rise to a peak near 40Hz in the power spectrum
Author: Dimitrios Bairaktaris
Medical Research Council (UK) Fellow,
Human Communication Research Centre,
University of Edinburgh,
2 Buccleuch Place, Edinburgh EH8 9LW, Scotland

Title: Multi-Layer Bidirectional Auto-Associative Memories

Abstract: This paper describes an auto-associative memory system, which is based on a combination of the Bidirectional Associative Memory (BAM) architecture and the use of randomly generated hidden representations called Randomized Internal Representations. The capacity and recall performance of the proposed system have been investigated under reduced connectivity and noise conditions. Analytical and simulation results suggested that the proposed system has an overall improved performance over other auto-associator networks. A multi-layer version of the basic system allows for storage and recall of non-random patterns. The system is capable of detecting novelty in the input when augmented with a plasticity-stability control mechanism. There is a number of functional and anatomical similarities between the proposed multi-layer system and the hippocampal formation in the brain.
Author: Harry Barrow and Alistair Bray
School of Cognitive and Computing Sciences
University of Sussex
Brighton, BN1 9QH, UK

Title: An Adaptive Neural Model of Early Visual Processing

Abstract: We describe an adaptive computer model of processing in the mammalian visual system, from retina to primary visual cortex, which incorporates: processing in retina and LGN; Non-negative neuron outputs and connection strengths; On-center and off-center channels; Excitatory and inhibitory cortical populations (3750 neurons); Non-linear cortical neurons, based on membrane potential and relaxation oscillator equations; Adaptive cortical connections. With images of the real world as input, the model robustly and simultaneously develops: oriented receptive field patterns, Gabor function-like fields, smoothly-varying orientation preference, and retinotopic mapping. Experiments with this and related models suggest the same mechanism may produce ocular dominance, and possibly color blobs
Author: Anthony J. Bell
AI-lab, Vrije Universiteit Brussel
Pleinlaan 2, B-1050 Brussels
BELGIUM

Title: Self-organising ion channel densities: the rationale for Anti-Hebb

Abstract: Although neural communication and integration are determined by the distribution of ion channels in the membrane, the role of voltage-dependent channels in neural integration is poorly understood. Here, an argument is presented for an `anti-Hebbian' rule which changes the distribution of intrinsic conductances in order to flatten voltage curvatures in dendrites. Simulations show that this rule can account for the self-organisation of dynamical receptive field properties such as resonance and direction selectivity, and also for faithful conduction of a signal in a cable. Various cellular mechanisms are proposed including activity-dependent migration of channel proteins in the plane of the membrane.
Author: Gregory S. Berns, Peter Dayan, Terrence J. Sejnowski
CNL, Salk Institute for Biological Studies,
PO Box 85800, San Diego, CA 92186-5800 USA

Title: Correlational-Based Development of Disparity Sensitivity

Abstract: A correlational-based model of development of disparity sensitivity is proposed. Two input layers are fully connected to a single cortical layer with fixed intra-cortical connections, and the weights modified by an unsupervised linear Hebb rule. A small amount of between-eye correlation (approximately 5% the amplitude of the same-eye correlation) leads to the development of binocular cortical cells with the left and right receptive fields aligned, i.e. zero disparity. Several cells which become monocularly dominated tend to have non-zero disparity preferences. The same relationship between ocular dominance and disparity has been experimentally observed in the visual cortex.
Author: E. K. Blum,
P. M. Khademi, D. G. Lavond, P. K. Leung & R. F. Thompson
Neural, Informational and Behavioral Sciences Program
University of Southem California, Los Angeles, 90089 CA.

Title: Modeling and Simulation of Cerebellar Networks Implicated in Classical Conditioning of the Rabbit Eyeblink Response

Abstract: Continuing previous research, several new detailed compartment models of cerebellar networks were developed based on new data from extra-cellular recordings in rabbit during conditioning of the eyeblink response. New sites of synaptic plasticity were postulated, as suggested by recent cooling-probe experiments, and these were incorporated into the models. The models were tested by simulation using the CAJAL simulator, which produces action potential spike trains that can be compared against the experimental time-trace data. Effects of cooling-induced lesions are simulated by varying certain parameters in the model neurons implicated by the lesion experiments. Conditioning is simulated by varying parameters in the model synapses of neurons implicated by these same experiments in a series of computer runs simulating training trials. Results provide further evidence that these cerebellar circuits may be the main neural substrates of this conditioning paradigm.
Author: Lyle J Borg-Graham and Norberto Grzywacz
Center for Biological Information Processing E25-201
Department of Brain and Cognitive Sciences
Massachusetts Institute of Technology
Cambridge, Massachusetts, 02139

Title: Non-Linear Spatio-Temporal Filtering in Dendrites: Directional Selectivity and Beyond

Abstract: We consider the non-linear spatio-temporal computation of directional selectivity (DS), and describe a model of retinal DS whose key elements are contained within the dendritic branches of amacrine cells. Interaction between excitatory and inhibitory synaptic input provides the non-linearity crucial for DS. Distal outputs, cable properties and synaptic kinetics underly spatial asymmetry, discrimination, and temporal delay, respectively. The model is supported by simulations of retinal cells and is consistent with recent electrophysiological data. Similar conditions may occur elsewhere in the brain, and simulations of hippocampal pyramidal cells suggest that dendritic filtering that is selective for synaptic activation sequence may be a more general phenomenon.
Author: James M. Bower and John H. Thompson

Title: A Silicon-based multi-single neuron probe designed for use in the cerebellum

Abstract: We have developed a silicon-based multichannel recording system for monitoring the single-unit activity of multiple neurons in the cerebellum. The system consists of a 15 channel multielectrode array, supporting electronics, and computers for data acquisition and analysis. The probe is a comb-like structure with 5 teeth and 3 recording sites per tooth. The geometry of the recording sites has been tailored to the anatomy of cerebellar cortex. Each tooth is very sharp and enters the brain with little or no dimpling. Recording sites, connecting leads, and wire-bonding pads are fabricated on a silicon substrate using thin-film technology. These probes have been used to record spontaneous granule cell, simple spike and complex spike responses in crus lla of the rat cerebellar hemisphere with excellent signal to noise characteristics.
Author: S.L. Bressler
Center for Complex Systems, Florida Atlantic U., Boca Raton, FL 33431
R.K. Nakamura
NIMH, Bethesda, MD 20892

Title: Inter-areal synchronization in rhesus macaque neocortex during a visual pattern discrimination task

Abstract: Functional relations between cortical sites were investigated in the behaving monkey. Transcortical field potentials were recorded from chronically implanted electrodes in macaques responding to one stimulus by lifting their hand from a lever within 500 msec (GO task), and to another by holding the lever (NO-GO task). FFTs of single-trial potentials were computed in a sliding window. Broad-band power increased during the response in the GO condition timelocked across cortical sites, and broad-band coherence between sites also increased. In individual trials, high-frequency oscillations briefly synchronized at the low-frequency amplitude peak.
Author: Paul C. Bush
Terrence J. Sejnowski
Howard Hughes
Medical Institute and Computational Neurobiology
Laboratory, Salk Institute, La Jolla, CA 92037, USA and University of
California at San Diego, La Jolla, CA 92093, USA.

Title: Simulations of Synaptic Integration in Neocortical Pyramidal Cells

Abstract: Despite their electrotonic compactness, neocortical pyramidal cells cannot be considered as point neurons because of nonlinear interactions between inputs on the same dendritic branch. Using compartmental simulations, we have shown that dendritic saturation is significant for physiological levels of synaptic activation. We have also confirmed earlier results showing that inhibition strong enough to produce a significant reduction in the input resistance of a cell does not prevent the firing of the cell if it is receiving strong excitatory input. Finally, we present a reduced pyramidal cell model (9 compartments) that runs significantly faster yet faithfully reproduces the behavior of the full 400 compartment model. The reduced model will be used for future physiological network simulations.
Author: Thomas C. Chimento,
David G. Doshayl, Muriel D. Ross
NASA-Ames, Biocomputation Center,
MS-239-11, Moffett Field, CA 94035
Steling Software, Palo Alto CA 94303

Title: Compartmental modeling of vestibular primary afferents: Ultrastructural morphology matters.

Abstract: Compartmental models were used to determine whether ultrastructural details of primary afferent neurons in the vestibular macula provide salient information for understanding cell function. Precise dimensions of primary afferent endings and the location of synapses were used as pararneters in computer simulations of passive current spread through the distal endings and action potential (AP) generation at the spike initiation zone (SIZ). It was demonstrated that synapse location and number, the morphology of the entire primary afferent and the morphology of collateral like processes had significant effects on depolarization magnitude and latency at the SIZ, thus deterrnining AP generation and latency. CNS*92 Abstract (50-100 words - For publication)
Author: A. D. Coop and S. J. Redman.
Division of Neuroscience, John Curtin School of Medical Research,
The Australian National University, GPO Box 334, 2601, Canberra, Australia.

Title: A Model of the Peristaltic Reflex

Abstract: A computational model of the peristaltic reflex containing ascending excitatory and descending inhibitory projections replicated much of the behaviour observed experimentally after reflex activation, including: i. orally directed contraction and anally directed relaxation of the circular muscle, ii. the induction of a propagating circular muscle contraction associated with bolus movement. Simulation results showed bolus movement to be: i. restricted by limiting the number of times sensory neurones could generate an action potential, ii. sensitive to its size and asymmetric weight distributions biased in favour of excitatory pathways, iii. maintained with overlapping projections only when the excitatory projections extended anally.
Author: Erik De Schutter and James M. Bower,
Div. of Biology 216-76, California
Institute of Technology, Pasadena CA 91125.

Title: Integration of synchronous and asynchronous synaptic inputs in a detailed
compartmental model of the cerebellar Purkinje cell.

Abstract: A detailed compartmental model of the Purkinje cell was used to examine integration of synchronous granule cell synaptic inputs during background asynchronous excitatory and inhibitory synaptic inputs. For a synchronous input on about 1% of the dendritic spines, the cell will fire on average one spike within 10 ms after the stimulus. This response is similar for inputs on proximal dendritic spines compared to inputs on distal spines. This insensitivity to the spatial location of synchronous inputs is caused by the P-type calcium channels in the dendritic membrane and by the branching pattern of the dendritic arbor.
Author: Virginia de Sat
desaQcs . rochester . edu
Computer Science Department
University of Rochester
Rochester, NY 14627-0226

Title: Self-teaching through correlated input

Abstract: How do cells in higher order sensory areas develop their invariant properties? In previous work we have shown that competitive learning coupled with a top-down teaching signal can produce compact invariant representations. In this paper we show that such a teaching signal can be derived internally from correlations between input patterns to two or more networks. Such correlations arise naturally from the structure present in natural environments. We demonstrate this process on two small but computationally difficult problems. We hypothesize that the correlations between and within sensory systems enable the learning of invariant properties.
Author: Peter F. Dominey
Hedco Neurosciences Building
University of Southern California
Los Angeles, CA, 90089-2520

Title: A Model of Cortex, Basal Ganglia and Thalamus in Stimulus-Response

Abstract: Novel visual stimuli are associated with saccadic eye movements. During learning, a visual stimulus is presented with a saccade target. Following a correct saccade, reward-related activation of the nigrostriatal dopamine system incrementally reinforces currently active visuo-striatal synapses. Striatum also receives input from FEF, and inhibits substantia nigra pars reticulata (SNr). SNr inhibits thalamic nuclei that project to FEF. Striatum assists activation of FEF via disinhibition of cortico-thalamic circuits. Presentation of a learned visual stimulus results in disinhibition of the thalamo-cortical circuit for the associated saccade. Background FEF activity is enhanced by the disinhibited thalamo-cortical circuit, resulting in the correct saccade.
Author: Dawei Dong
Lawrence Berkeley Laboratory,
University of California MS 70A-3307,
1 Cyclotron Road, Berkeley, California 94720

Title: Dynamics of hebbian learning in feedback networks and development of interconnection within visual cortex

Abstract: Two kinds of dynamic processes take place in neural networks. One involves the change with time of the activity of each neuron. The other involves the change in strength of the connections (synapses) between neurons. When a neural network is learning or developing, both processes simultaneously take place, and their dynamics interact. This interaction is particularly important in feedback networks. Theoretical framework is developed to help understand the combined activity and synapse dynamics for a class of such adaptive networks. The methods and viewpoint are illustrated by using them to describe the development of orientation selective cells in cat primary visual cortex. Within this model, orientation selectivity originates from feedback pathways within an area of cortex, rather than feedforward pathways between areas.
Author: Georg Dorffner
Thomas Schonauer
Dept.of Medical Cybernetics and Artificial Intelligence,
University of Vienna,
and Austrian Research Institute for Artificial Intelligence

Title: Unsupervised Learning of Simple Speech Production Based on Soft Categorization

Abstract: In this paper a novel approach to the self-organization of a sensory-motor loop is proposed. A neural network receiving spectra of stationary signals as input and producing the motor parameters for an articulatory model is trained to reproduce perceived vowels. Learning consists of two components. First, categorization in alayer that gradually learns to compress inputs into more and more unitized responses (soft competitive learning). Secondly, adaptation of the weights between this and the output layer, taking as learning rate a measure of goodness-of-fit easily computed for each response in the categorization layer. The model learns by exploring the possible articulatory space - random, at first, then increasingly guided by the categorical repsonses - and adpating its weights when a signal, produced by itself, has a large goodness-of-fit with one of the learned categories.
Author: David Egert,
U.C. Berkeley/U.C. San Francisco
Graduate Group in Bioengineering

Title: Hair-Cell Modelling to Explore the Physiological Basis of Tuning in the Lower Vertebrate Ear

Abstract: The network modeling program SPICE was used to realize a model of hair-cell membrane dynamics and their role in the tuning processes of lower vertebrates. Driving the modeled system with sinusoidal strain, we observed the transfer relationship between strain and membrane potential. The observed transfer relation was of low dynamic order and was not consistent with the transfer relations we have seen in intact end-organs of the bullfrog. Hence we suspect the tuning involves interactions between several hair-cells, perhaps facilitated by mechanical linkages. (Work supported by NIH grant DC00112.)
Author: P. Erdi,
Biophysics Group,
Central Research
Institute for Physics,
Hungarian Academy of Sciences,
H-1525 Budapest P.O. Box 49, Hungary

Title: Dynamic phenomenia in the olfactory bulb

Abstract: A mathematical model of the olfactory bulb is presented to study the dynamics of the bulbar information processing. A two level model is adopted to describe both neural activity and synaptic modifiability. The model takes explicitly into account the existence of lateral interactions in the mitral layer, and the synaptic modifiability of these connections. A series of bifurcation phenomena among fix points, limit cycle and strange attractors have been demonstrated. Chaos occurred only in the case of excitatory lateral conncetions. Coexistence between oscillation and chaos, and synaptic modification induced transition have also been found.
Author: E. Erwin,
K. Obermayer and K. Schulten
Beckman Institute and Department of Physics
University of Illinois at Urbana-Champaign
Urbana, IL61801,U.S.A.

Title: A Critical Comparison of Models of Cortical Visual Map Formation

Abstract: High-resolution images of orientation and ocular dominance columns in monkey striate cortex have recently been obtained by optical techniques, allowing a quantitative description of the spatial patterns on the local and global scales. The new data has provided strong evidence that the patterns of orientation and ocular dominance in monkey striate cortex are not independent, but rather correlated locally. In our contribution we compare the predictions of several models of cortical map formation, in light of the new data. The comparison is made in terms of Fourier transforms, auto-correlation functions, Gabor transforms, and the gradients of the spatial patterns. Our investigation includes topology-preserving maps and elastic-net models, models of correlation-based learning, and non-developmental pattern models. When possible, the models are also extended to account for the simultaneous formation of orientation and ocular dominance columns, and hence to allow predictions about correlations between the two maps.
Author: L. J. Feinswog,
R. K. Hutson, G. T. Kenyon and D. C. Tam
Division of Neuroscience
Baylor College of Medicine
Houston, TX 77030

Title: An Object-Oriented Paradigm for the Design of Realistic Neural Simulators

Abstract: We implement a detailed, realistic neural simulation using an objectoriented programming paradigm incorporating the known properties of biological neurons. The object-oriented design eliminates many of the complexities typical of a conventional modeling approach. Our model allows implementation of multiple integration algorithms which can be selected interchangeably by the user. Descriptions of physiological objects are supported at several hierarchical levels, from networks to neurons, membrane compartments, and ionic channels. Our design provides the flexibility to incorporate new physiological features and numerical algorithms with fault-tolerant capabilities.
Author: Leif H. Finkel and Paul Sajda
Department of Bioengineering and
Institute of Neurological Sciences
University of Pennsylvania
220 South 33rd Street
Philadelphia, PA 19104-6392

Title: Computer Simulations of Object Discrimination by Visual Cortex

Abstract: We present computer simulations of how the visual cortex may discriminate objects based on depth-from-occlusion. We propose neural mechanisms for how the visual system binds edges into contours, and binds contours and surfaces into objects. The model is simulated by a system of physiologically-based neural networks which feature feedback connections from higher to lower cortical areas, a distributed representation of depth, and phase-locked cortical neuronal firing. The system demonstrates psychophysical properties consistent with human perception of real and illusory visual scenes. The model addresses both the binding problem and the problem of object segmentation.
Author: Peter Foldiak
MRC Research Centre In Brain and Behaviour,
Department of Experimental Psychology,
University of Oxford,
South Parks Road,
Oxford OX1 3UD, U.K.

Title: The 'Ideal Homunculus': statistical inference from neural population responses

Abstract: An 'ideal homunculus' makes inference about the stimulus based on the responses of a group of neurons. The tuning curves of individual neurons can be interpreted as a conditional probability distribution (response | stimulus). The Bayes rule can be used to calculate the distribution of possible stimuli given the recorded response, (stimulus | response). Combining responses from more than one neuron 'sharpens' the distribution, and the information content of the individual contributions can be calculated. The homunculus can be looking at groups of neurons at different levels; an example is given where orientation is inferred from responses of V1 neurons.
Author: William R. Foster, J.F.R. Paton, James S. Schwaber, Lyle H. Ungar
Department of Chemical Engineering, 311A Towne Bldg.,
University of Pennsylvania,
Philadelphia, Pa 19104, U.S.A.

Title: Matching neural models to experiment

Abstract: Methods of quantitatively fitting single compartment neuron models to experimentally observed neuron behavior in current clamp are investigated. Standard minimization techniques relying on gradient information are not useful for matching Hodgkin and Huxley style models to current clamp data. It is, however, possible to fit current clamp data with minimal experimental information by making use of channel kinetics descriptions taken from the literature. The fitting procedure uses random search and an appropriately defined metric. This approach eases the task of creating neuron models for use in network simulations. Random search may also be used to operate on simplified channel kinetics in order to create accurate matches with experimental data. The resulting fits to experiment are not unique and therefore the utility of the models must be judged in terms of their ability to represent important aspects of observed neuronal behavior over wide ranges of inputs.
Author: Erik Fransen,
Anders Lansner and Hans Liljenstrom
SANS - Studies of Artificial Neural Systems
Dept. of Numerical Analysis and Computing Science
Royal Institute of Technology,
S-100 44 Stockholm, Sweden

Title: A Model of Cortical Associative Memory based on Hebbian Cell Assemblies

Abstract: A model of cortical associative memory, based on Hebb's theory of cell assemblies, has been developed and simulated. The network is comprised of realistically modelled pyramidal-type cells and inhibitory fast spiking interneurons and its connectivity is adopted from a trained recurrent artificial neural network. After-activity, pattern completion and competition between cell assemblies is readily produced. If, instead of pyramidal cells, motor neurons are used, spike synchronization can be observed but after-activity is hard to produce. After-activity is facilitated by increased levels of serotonin and disrupted by low levels. Our results support the biological plausibility of Hebb's cell assembly theory.
Author: Algis Garliauskas, Algirdas SHIMOLIUNAS
Institute of Mathematics and Informatics,
Lithuanian Academy of Sciences,
2600, Vilnius, Akademijos St.4, Lithuania

Title: Logic functions realized on a stationary nonlinear dendrite

Abstract: The binary logic functions "AND" and "AND/OR" are realized by the model of a nonlinear stationary dendritic branch. The neuron with such dendrites is a complex logic system performing a great number of elementary logicoperations.
Author: B.P. Graham,
Centre for Information Science Research
Prof. S.J. Redman, Division of Neuroscience,
The John Curtin School of Medical Research
The Australian National University, Canberra

Title: Simulation of the Muscle Stretch Reflex by a Neuronal Network

Abstract: The roles that a variety of different neuronal types play in the muscle stretch reflex have been investigated by means of a dynamic computer simulation. The simulated reflex circuit controlled two antagonistic muscles acting on a load. Simulation results revealed that while the dynamic component of the Group 1a afferent firing rate was important for good dynamic response, it also caused oscillations in the load position at a frequency similar to the condition known as clonus. If Group II afferents also excited the motoneurones, these oscillations were reduced. Inhibition from Renshaw cells and Group 1b afferents greatly increased the stability of the circuit.
Author: By E.B. Graves, D.W. RICHTER* AND J.S.SCHWABERl
Neural Computation Group, DuPont Experimental Station E352,
Dupont Company,
Wilmington, DE 19880-0352,
U.S.A. and *Physiology Institute,
Georg-August-University of Goettingen,
Humboldtallee 23, Goettingen, FRG

Title: Network Model of the Respiratory Rhythm

Abstract: We propose a model for the generation of the respiratory rhythm. The model is composed of four distinct neuron types recorded in vivo in the cat, defined by their phase relationship to the respiratory cycle as seen in the phrenic neurogram. It is proposed that the respiratory rhythm emerges from the connectivity and membrane dynamics of these neurons. Computational models are created of the four cell types using Hodgkin-Huxley form kinetics to describe active membrane properties that are then tuned to reproduce the recorded patterns of neuronal behavior. The neurons are reciprocally connected with inhibitory synapses. Simulation results accurately reproduce the three-phased respiratory cycle, and demonstrate the way in which this activity may arise from intrinsic membrane, as well as network, properties.
Author: Richard A. Gray
Department of Biology Gilmer Hall
University of Virginia,
Charlottesville, VA 22901
second author:
W. Otto Friesen
Department of Biology
Gilmer Hall

Title:

Abstract: We have developed a technique to determine the natural frequencies of biological oscillators and have shown that these frequencies have physiological significance. We passed both sub-threshold and supra-threshold sinusoidal current of several frequencies into the Hodgkin-Huxley mathematical model of the squid axon (Hodgkin and Huxley 1952). The resonance peaks for the resulting membrane potential associated with these two experiments occurred at 68 Hz and 174 Hz. These frequencies compare favorably to the range of firing frequencies (68 Hz to 169 Hz) of the Hodgkin-Huxley model response to constant current. We have also used this resonance technique to find a resonance peak at approximately 1 Hz in the membrane potential of a leech motor neuron which agrees with the normal swimming frequency of the leech (Pearce and Friesen 1985, Kristan et al. 1974). correspondence should be addressed to: