Energy-efficient neuromorphic computing

Researchers report deep convolutional neural networks, which are biologically inspired adaptive computing programs, designed to perform visual and auditory classification tasks on neuromorphic hardware, a platform for running neural networks, at more than 6,000 frames per second per Watt, a necessary step toward fast, efficient, human-level, brain-inspired computing.

Article #16-04850: “Convolutional networks for fast, energy-efficient neuromorphic computing,” by Steven K. Esser et al.