A Neuromorphic Architecture Evaluation Platform with FPGAs

Neuromorphic computers offer tremendous opportunities for the future of computing. However, few deployment and application-ready neuromorphic hardware systems have been developed.  Therefore, the design of neuromorphic computer architectures and the effect of design choices on processing speed, power, and accuracy are not well understood.  To address this issue, we propose to develop a neuromorphic architecture codesign and evaluation platform using software simulation and field programmable gate arrays (FPGAs) to understand the processing speed, power, and accuracy ramifications of potential neuromorphic architecture designs.  Our goal is to provide a platform that evaluates the software and hardware effects of new neuromorphic architectures and to provide a baseline of hardware performance for future custom chip implementations that uses alternative devices and materials.