COMPUTATIONAL SYMBOLIC MODELS OF NEURAL NETWORK DYNAMICS

这项研究的长期目标是创造一种动态的、象征性的、

The long-term goal of this research is to produce a dynamic, symbolic, computer simulation of a functioning biological neural network. The purpose is to use the simulation as an investigative tool for studying factors contributing to neural system robustness, plasticity, health and disease. The simulation will be based on mathematical and physical interpretations of actual neural architecture and electrophysiology in a mammalian neural network, the vestibular macula, which has many of the attributes of more advanced systems. The proposed investigation will fill the present gap in our knowledge of the relationship between network geometry and neural coding and will improve upon an existing simulation, making it a better research tool. The specific aims of this study are to 1) identify electrophysiologically distinct classes of vestibular afferents in gerbils and to label them intracellularly with horseradish peroxidase (HRP), 2) reconstruct the labeled nerve/terminal receptive fields, synapses, efferent boutons, and surrounding portions of the neural network as three dimensional solid images, 3) evaluate the correspondence between functional clailsification and vestibular neural geometry and synaptology using a compartmental model, 4) develop algorithms to improve a current simulation so that it more closely mimics vestibular network dynamics. We propose a ten-compartment model that has anatomically defined boundaries emphasizing synaptic zones and branch confluences. The compartmental model will be incorporated into a symbolic, dynamic simulation that will be used to determine how neural geometry and the immediate network environment influence nerve discharge patterns. Some experiments will test for properties that might underlie system robustness, such as redundancy and constrained randomness in wiring; others will explore the influence of intrinsic feedforward-feedback loops and of extrinsic, presynaptic modulation on neural discharge patterns and coding. The model can be easily manipulated to study the contributions of component parts to network dynamics, health and disease. The simulation is directly applicable to the study of some inner ear disorders resulting in dysequilibrium, such as Meniere's syndrome, or the study of the effects of ototoxic drugs, by selective omission of components such as type I hair cells. The simulation has broader significance as a model for studying normal and abnormal functioning of more advanced neural networks. Many features of the macular neural network are shared by more complex biological systems: for example, non-modularity of processing elements, feedforward-feedback loops, segmentation, synaptic weighting, and reciprocal synapses. Related work will apply nonlinear dynamical systems theory to nerve pulse trains recorded during electrophysiological testing, in order to learn the role of chaos in neural health and in neural plasticity.