An Analog/Digital Simulation System for Biomimetic Neural Networks
Résumé
This work presents a tool addressing the investigation of spiking neural networks, in a tentative understanding of the temporal coding of information by such networks. Detailed and biologically-realistic models of neurons can then be implemented to simulate neural dynamics at the single cell level. Network-level simulations are also possible, including the computation of adaptive functions (connectivity and plasticity rules). We will present a system we engineered to allow real-time simulations of neural networks. It combines analog VLSI artificial neurons with software modules, with enough flexibility on the models to allow a systematic exploration of the neural network dynamics, including adaptation rules.