Self-Knowledge Distillation via Dropout

Hyoje Lee, Yeachan Park, Hyun Seo +1

To boost the performance, deep neural networks require deeper or wider network structures that involve massive computational and memory costs. To alleviate this issue, the self-kno…