Latent Weight Quantization for Integerized Training of Deep Neural Networks | IEEE Journals & Magazine | IEEE Xplore