Learning without Forgetting for Decentralized Neural Nets with Low Communication Overhead | IEEE Conference Publication | IEEE Xplore