Coupled Neural-Dendritic Processes: Cooperative Stochastic Effects and the Analysis of Spike Trains.

摘要

We can create a richer and more neurophysiologically realistic model of neural activity in the brain by developing a model of neural-dendritic coupling, one which expressly accounts for the way in which the many afferent connections into the neural body influence the somatic membrane potential. Such a model would begin to fill the need within the Artificial Neural Network community for neural models which go beyond the current weighted sum paradigm for artificial neuron connectivity. Although such models have use in engineering applications, there are many aspects of biological neural-dendritic organization which could enrich artificial neural networks. Moving from simple axonal connection weight neural models to neural-dendritic models with a richer structure will allow investigation of both events at the neural level (e.g., inter-spike interval histograms and stochastic resonance) and also potentially at the neural systems level. This will also introduce the possibility of introducing cross-scale interactions into artificial neural systems.