A simple model of spike processing

We describe a simple model of spike processing build with a number of neural hardware primitives including integrate-and-fire neurons, dendritic trees, integrators, inhibition logic and one-to-many axonal/dendritic tree connectivity. Functionally, our model consists of neuro-modulators, communication channels, neuro-demodulators and filters. Integrate-and-fire neurons play the role of neuro-modulators. They represent dendritic currents in the spike domain through a process of reversible computation. Neuro-demodulation and irreversible computation takes place in dendritic trees as time domain linear and/or non-linear operations. Assuming that the stimulus at the input of an integrate-and-fire sensory neuron is bandlimited, we demonstrate how to construct a linear operator that maps an arbitrary stimulus into a desired neuronal signal. (c) 2006 Elsevier B.V. All rights reserved.