Regresar

Synfire chain emulation by means of flexible SNN modeling on a SIMD multicore architecture

Abstract:

The implementation of a synfire chain (SFC) application that performs synchronous alignment mapped on a hardware multiprocessor architecture (SNAVA) is reported. This demonstrates a flexible SNN modeling capability of the architecture. The neural algorithm is executed by means of a digital Spiking Neural Network (SNN) emulator, using single instruction multiple data (SIMD) processing. The flexibility and capability of SNAVA to solve complex nonlinear algorithm was verified using time slot emulation on a customized neural topology. The SFC application has been implemented on an FPGA Kintex 7 using a network of 200 neurons with 7500 synaptic connections.