Energy-efficient neural networks using approximate computation reuse | IEEE Conference Publication | IEEE Xplore