arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2106.04570cs.LGcs.AIcs.CLcs.CV
收藏

BERT学会教学:元学习与知识蒸馏

BERT Learns to Teach: Knowledge Distillation with Meta Learning

Stanford University(斯坦福大学);University of California, San Diego(加州大学圣地亚哥分校)

Wangchunshu Zhou, Canwen Xu, Julian McAuley

摘要:

我们提出了知识蒸馏与元学习(MetaDistil),一个简单而有效的替代传统的知识蒸馏(KD)的方法,教师模型是固定的培训。我们证明了教师网络可以学习更好地将知识传递给学生网络(即,学习教学),并具有来自元学习框架中的提炼学生网络的性能的反馈。此外,我们引入了一个试点更新机制,以提高对齐的内部学习者和元学习者之间的元学习算法,专注于一个改进的内部学习者。在各种基准测试上的实验表明,MetaDistil与传统KD算法相比,可以产生显著的改进,并且对不同学生容量和超参数的选择不太敏感,便于在不同任务和模型上使用KD。

英文摘要:

We present Knowledge Distillation with Meta Learning (MetaDistil), a simple yet effective alternative to traditional knowledge distillation (KD) methods where the teacher model is fixed during training. We show the teacher network can learn to better transfer knowledge to the student network (i.e., learning to teach) with the feedback from the performance of the distilled student network in a meta learning framework. Moreover, we introduce a pilot update mechanism to improve the alignment between the inner-learner and meta-learner in meta learning algorithms that focus on an improved inner-learner. Experiments on various benchmarks show that MetaDistil can yield significant improvements compared with traditional KD algorithms and is less sensitive to the choice of different student capacity and hyperparameters, facilitating the use of KD on different tasks and models.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑