Simulation of neural networks: a preliminary report on two computational strategies.

Abstract

The basic assumptions of the present contribution are the following: i) a satisfactory mechanistic knowledge of the higher human cognitive abilities will be reached after understanding the concerted and cooperative behavior of brain regions usually identified by peculiar morphofunctional features; ii) it is really hard to overestimate the importance of computer-aided modeling and simulation in the study of the global and local network(s) connecting the above regions; iii) the related computational problems are better faced by exploiting different software tools, each endowed with excellent performance in specific problems, than focussing on a single programming environment. In this frame, we tested the steepness of the learning curve of two popular neural simulation environments, namely NEURON and BRIAN, in modeling the following cases: a) a selfsustained, reciprocal activation of a small number of ring-chained neurons, and b) the spiking activity of a small network possibly endowed with a random set of connections. We confirm that the NEURON and the BRIAN environments appear well suited in dealing with case a) and b), respectively, and, in particular, in studying the biophysical / morphological features of single neural cells or in exploring the functional features even of complicated or large-size network topologies.