There could be a new era of codesign dawning for machine learning, one that moves away from the training and inference separations and toward far less dense networks with highly sparse weights and ...