A simple non-synaptic memory mechanism

s / Neuroscience Research 68S (2010) e109–e222 e213 brane currents in the population. The neuronal populations are designed to have realistic cell densities and geometric arrangements, and contributions to the LFP at different recording positions are calculated. Each neuron receives numerous synaptic inputs with tailored presynaptic spike-train patterns. This enables us to investigate how the population geometry and the statistics of the presynaptic spike trains such as spike-train correlations and spectral properties determine the spatial range of the LFP. doi:10.1016/j.neures.2010.07.2510 P1-r01 Input dependent cell assembly dynamics in an inhibitory spiking network model Adam P.D. Ponzi , Jeff Wickens Neurobiology Research Unit, Okinawa Institute of Science and Technology We present a model of an inhibitory spiking network composed of striatal medium spiny neurons (MSN). We show that the network exhibits complex dynamics produced by the occurrence of episodically firing cell assemblies, and that this occurs for various choices of MSN cell model and is thus a network generated property, while details of cell spiking statistics do depend on the cell model. When excitatory input to the network is dynamically variable we show cell assemblies display input onset dependent dynamics even when the excitatory inputs to each cell are composed of many Poisson spike trains with variable rates, providing the excitatory synapses obey certain conditions. We discuss how this behaviour may relate to striatal cognitive functions in behavioural tasks. doi:10.1016/j.neures.2010.07.2511 P1-r02 A simple non-synaptic memory mechanism S. Shuichi Haupt 1 , Tomoki Kazawa 1, Ikuko Nishikawa 2, Ryohei Kanzaki 1 1 RCAST, Intelligent Cooperative Systems, University of Tokyo 2 College of Information Science and Engineering, Ritsumeikan University, Kusatsu, Shiga 525-8577, Japan While memory storage in neural networks is most commonly associated with changes in synaptic weights, other forms of memory are feasible. For example, Lisman & Idiart (Science 267: 1512,1995) have shown that a network can store memories using nested oscillations of different frequency bands. We sought to implement a very primitive mechanism for storing one bit by switching between two states of a 2-cell network. A coupled oscillator model in which two neurons are connected by a gap junction has been investigated by Cymbalyuk et al (Biol Cybern 71: 153, 1994). The system receives a stationary bias current input that can induce steady, oscillatory, and chaotic states. Of particular interest is a regime in which the bias input (constant neuromodulatory excitation), can induce two types of non-symmetrical limit cycles in which oscillation amplitudes of the neurons differ. Switching between such limit cycles could be a reliable way to store a single bit. This is possible by disturbing the system sufficiently and letting it settle back into a stable (nonsymmetrical) limit cycle. Conditions for reliable, phase-independent state switching consist of input pulses over a certain range of durations, which are strongly imbalanced between the two coupled cells: an excitatory input for one cell and an ihibitory input for the other cell occurring simultaneously. An application of this system is the flip-flop mechanism in the silkmoth brain involved in pheromone orientation behaviour in which neurons switch between low and high states upon pheromone stimulation like a toggle flipflop. Assuming the output of the two neurons of the coupled oscillator to have opposite sign and both outputs being integrated by a flip-flop neuron, the neuron will be inhibited when the coupled oscillator is in the state in which the inhibitory cell’s oscillation amplitude is higher and excited when the oscillator is in the alternative state. doi:10.1016/j.neures.2010.07.2512 P1-r03 Olfactory center model of land slug using pulsetype hardware chaotic neuron models Ken Saito 1 , Yu-ta Hamasaki 2, Hirokazu Hatano 2, Minoru Saito 2, Fumio Uchikoba 1, Yoshifumi Sekine 3 1 Department of Precision Machinery Engineering, College of Science and Technology, Nihon University, Chiba, Japan 2 Graduate School of Integrated Basic Sciences, Nihon University, Tokyo, Japan 3 Department of Electronics and computer Science, College of Science and Technology, Nihon University, Chiba, Japan Electrical oscillatory activities are a ubiquitous feature in nervous systems. The oscillatory patterns play an important role in the processing of sensory information recognition. For example, earlier reports describe that the oscillatory patterns in the olfactory center (procerebrum; PC) of the land slug Limax are changed by odor stimuli to the tentacles. Previously, we examined the odor responses of the local field potential (LFP) in the PC by extracellular recording and showed that the reconstructed attractors were different between the LFP oscillations before and after the odor stimulus. Olfactory processing has also been studied in relation to rabbits and land slugs through the construction and use of mathematical neural network models. However, a large-scale model is necessary for the construction of a model which realizes sensory information recognition by the oscillatory patterns. Therefore, a hardware model which can generate the oscillatory patterns is desired because nonlinear operations can be processed at higher speeds than the mathematical model. We are studying about the neural network using hardware neuron models to construct the olfactory center model of the living organisms. In the present study, we discuss about the oscillatory pattern generation using pulse-type hardware chaotic neuron models. Our model shows periodic, quasi-periodic and chaotic oscillations similar to those of the PC of the land slug by changing the synaptic connection weights or the external pulse stimuli and the attractors can be different between the oscillation patterns. doi:10.1016/j.neures.2010.07.2513 P1-r04 Neuromorphic CMOS analog circuit exhibiting array-enhanced stochastic resonance Behavior with population heterogeneity Tovar Gessyca , Tetsuya Asai, Yoshihito Amemiya Graduate school of Information science and technology, Hokkaido University. Sapporo Stochastic Resonance (SR) is a phenomenon in which the response of a system can be enhanced in the presence of an optimal level of noise. It is well known that there are several noise sources in the nervous system and neurons are subject to these noises. Moreover, it has been observed that SR gets further enhanced by local couplings between neurons (array-enhanced SR). For example, Stacey et al. demonstrated that an array of simulated hippocampal CA1 neurons exhibited SR-like behaviors where the correlation between the output and a sub-threshold input signal in the network increased as the coupling strength between the neurons was increased, and the correlation value was further increased as the number of the neurons increased. Motivated by these findings, we proposed a neural network model composed of Wilson-Cowan neuron models that is suitable for neuromorphic semiconductor devices . In the network each neuron device is electrically coupled to its four neighbors to form a 2D grid network. All the neurons accept a common sub-threshold input. The firing of each neuron is recorded and is converted into a series of pulses of amplitude 1 and 0 corresponding to the firing and non-firing states respectively. The output of the network is then defined by the sum of all the pulses. We carried out circuit simulations using a simulation program with integrated circuit emphasis (SPICE), with typical CMOS device parameters, and confirmed that without the electrical coupling, the circuit network exhibited a standard SR behavior and that the correlation value was further increased as the coupling strength increased. These results indicate that, as suggested by biological and simulated evidence, unreliable (possible semiconductor nano) devices could be use to develop reliable information processing systems. doi:10.1016/j.neures.2010.07.2514