A Hardware Accelerated Simulation Environment for Spiking Neural Networks

Spiking Neural Networks (SNNs) model the biological functions of the human brain enabling neuro/computer scientists to investigate how arrays of neurons can be used to solve computational tasks. However, as network models approach the biological scale with significantly large numbers of neurons, existing software simulation environments face the problem of scalability and increasing simulation times. Emulation in hardware offers a significant increase in the acceleration of simulations through the exploitation of parallelism and dedicated on-chip training. However, it is important that the configuration of hardware emulation is abstracted from the novice end-user to allow flexible, high-level specification and execution. This paper presents a novel reconfigurable hardware architecture and internet-based configuration environment for the FPGA-based acceleration of SNNs with online training. Results are presented to demonstrate the acceleration performance.

• Type: Conference
• Year: 2009
• Where: ARC
• Authors: Brendan P. Glackin, Jim Harkin, T. Martin McGinnity, Liam P. Maguire
• Keywords: Acceleration, ARC 2009, Brain Enabling Neuro/computer, FPGA-based Acceleration, Hardware