arXivDaily arXiv每日学术速递 周一至周五更新
北美退伍军人节(Veterans Day),12 日无更新
arXiv 2005.06537cs.CL
收藏

混合使用$h-1$股骨头优于$h$股骨头

A Mixture of $h-1$ Heads is Better than $h$ Heads

University of Washington(华盛顿大学);Allen Institute for Artificial Intelligence(艾伦人工智能研究所)

Hao Peng, Roy Schwartz, Dianqi Li, Noah A. Smith

摘要:

多头注意神经架构在各种自然语言处理任务上取得了最先进的成果。有证据表明,它们是过度参数化的;注意力头可以被修剪,而不会造成显着的性能损失。在这项工作中,我们相反地“重新分配”它们-模型学习在不同的输入上激活不同的头部。将多头注意力与混合专家模型联系起来,提出了混合注意专家模型(MAE)。MAE使用块坐标下降算法进行训练,该算法在更新(1)专家的职责和(2)他们的参数之间交替。在机器翻译和语言建模上的实验表明,MAE在这两个任务上都优于强基线。特别是,在WMT 14英语到德语的翻译数据集上,MAE比“transformer-base”提高了0.8 BLEU,参数数量相当。我们的分析表明,我们的模型学习将不同的专家专门化到不同的输入。

英文摘要:

Multi-head attentive neural architectures have achieved state-of-the-art results on a variety of natural language processing tasks. Evidence has shown that they are overparameterized; attention heads can be pruned without significant performance loss. In this work, we instead "reallocate" them -- the model learns to activate different heads on different inputs. Drawing connections between multi-head attention and mixture of experts, we propose the mixture of attentive experts model (MAE). MAE is trained using a block coordinate descent algorithm that alternates between updating (1) the responsibilities of the experts and (2) their parameters. Experiments on machine translation and language modeling show that MAE outperforms strong baselines on both tasks. Particularly, on the WMT14 English to German translation dataset, MAE improves over "transformer-base" by 0.8 BLEU, with a comparable number of parameters. Our analysis shows that our model learns to specialize different experts to different inputs.

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。

↑