Johns Hopkins electrical and computer engineers are pioneering a new approach to creating neural network chips—neuromorphic accelerators that could power energy-efficient, real-time machine ...