Long Short-Term Sample Distillation

Liang Jiang, Zujie Wen, Zhongping Liang +7

In the past decade, there has been substantial progress at training increasingly deep neural networks. Recent advances within the teacher--student training paradigm have establishe…