1 paper
Weipeng Huang, Junjie Tao, Changbo Deng +4
This paper revisits Deep Mutual Learning (DML), a simple yet effective computing paradigm. We propose using Rényi divergence instead of the KL divergence, which is more flexible a…