Parametric Architecture for Modeling Neuronal Systems

This paper presents a new approach to the neural systems modeling under the scope of function evaluation. The proposal argues for a computational framework based on the use of a primitive operation, namely a weighted sum defined by a two input table, which converts the calculation of function values into a recursive operation. The implementation of this approach is a simple circuit which can be tuned to perform the function calculation by only changing the parameter values of the primitive stored in the table. The calculation of the values is carried out by the successive iterations of the primitive. The application of this approach is concerned with the simulation of the behavior of some neural structures that exhibit different bursting patterns. We consider the bursts as concatenated function fragments that can be simulated with a satisfying accuracy by our method. When compared with other well-known proposals based on biophysical or mathematical modeling, our approach provides a satisfying trade-off between area costs and time delay.