SqueezeBlock: A Transparent Weight Compression Scheme for Deep Neural Networks | IEEE Conference Publication | IEEE Xplore