wangqiangneu / MT-PaperReading

Record my paper reading about Machine Translation and other related works.
36 stars 2 forks source link

20-ACL-Hard-Coded Gaussian Attention for Neural Machine Translation #63

Open wangqiangneu opened 4 years ago

wangqiangneu commented 4 years ago

简介

multi-head self attention下手,试图简化。方法很简单,用no parameter的hard-code gaussian distribution做,跟之前的local attention差不多,只不过mean不用参数学,直接指定好了, e.g. (i-1, i, i+1)似乎mean的设定还有挺大影响,不是简单的i就完事了。方法很简单,没啥说的,主要是实验分析那块做了比较多

有意思的点

论文信息

总结