Autor:	Shazeer, Noam, Lan, Zhenzhong, Cheng, Youlong, Ding, Nan, Hou, Le
Rok vydání:	2020
Předmět:	Computer Science - Machine Learning Computer Science - Neural and Evolutionary Computing Computer Science - Sound Electrical Engineering and Systems Science - Audio and Speech Processing Statistics - Machine Learning
Druh dokumentu:	Working Paper
Popis:	We introduce "talking-heads attention" - a variation on multi-head attention which includes linearprojections across the attention-heads dimension, immediately before and after the softmax operation.While inserting only a small number of additional parameters and a moderate amount of additionalcomputation, talking-heads attention leads to better perplexities on masked language modeling tasks, aswell as better quality when transfer-learning to language comprehension and question answering tasks.
Databáze:	arXiv
Externí odkaz:	http://arxiv.org/abs/2003.02436 Zobrazit plný text záznamu View this record from Arxiv