Neural network training could one day require less computing power and hardware, thanks to a new nanodevice that can run neural network computations using 100 to 1000 times less energy and area than existing CMOS-based hardware.
Artificial neuron device could shrink energy use and size of neural network hardware
