Publication

Equivalent-accuracy accelerated neural-network training using analogue memory